Transport Layer 3-1
Chapter 3Transport Layer
Computer NetworkingA Top Down Approach5th editionJim Kurose Keith RossAddison-Wesley April2009
All material copyright 1996-2009JF Kurose and KW Ross All Rights Reserved
Transport Layer 3-2
Chapter 3 Transport LayerOur goals understand principles
behind transportlayer services multiplexingdemultipl
exing reliable data transfer flow control congestion control
learn about transportlayer protocols in theInternet UDP connectionless
transport TCP connection-oriented
transport TCP congestion control
Transport Layer 3-3
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-4
Transport services and protocols provide logical communication
between app processesrunning on different hosts
transport protocols run in endsystems send side breaks app
messages into segmentspasses to network layer
rcv side reassemblessegments into messagespasses to app layer
more than one transportprotocol available to apps Internet TCP and UDP
applicationtransportnetworkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-5
Transport vs network layer
network layer logicalcommunicationbetween hosts
transport layer logicalcommunicationbetween processes relies on enhances
network layer services
Household analogy12 kids sending letters
to 12 kids processes = kids app messages = letters
in envelopes hosts = houses transport protocol =
Ann and Bill network-layer protocol
= postal service
Transport Layer 3-6
Internet transport-layer protocols
reliable in-orderdelivery (TCP) congestion control flow control connection setup
unreliable unordereddelivery UDP no-frills extension of
ldquobest-effortrdquo IP services not available
delay guarantees bandwidth guarantees
applicationtransportnetworkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-7
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-8
Multiplexingdemultiplexing
application
transport
network
link
physical
P1 application
transport
network
link
physical
application
transport
network
link
physical
P2P3 P4P1
host 1 host 2 host 3
= process= socket
delivering received segmentsto correct socket
Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)
Multiplexing at send host
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-2
Chapter 3 Transport LayerOur goals understand principles
behind transportlayer services multiplexingdemultipl
exing reliable data transfer flow control congestion control
learn about transportlayer protocols in theInternet UDP connectionless
transport TCP connection-oriented
transport TCP congestion control
Transport Layer 3-3
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-4
Transport services and protocols provide logical communication
between app processesrunning on different hosts
transport protocols run in endsystems send side breaks app
messages into segmentspasses to network layer
rcv side reassemblessegments into messagespasses to app layer
more than one transportprotocol available to apps Internet TCP and UDP
applicationtransportnetworkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-5
Transport vs network layer
network layer logicalcommunicationbetween hosts
transport layer logicalcommunicationbetween processes relies on enhances
network layer services
Household analogy12 kids sending letters
to 12 kids processes = kids app messages = letters
in envelopes hosts = houses transport protocol =
Ann and Bill network-layer protocol
= postal service
Transport Layer 3-6
Internet transport-layer protocols
reliable in-orderdelivery (TCP) congestion control flow control connection setup
unreliable unordereddelivery UDP no-frills extension of
ldquobest-effortrdquo IP services not available
delay guarantees bandwidth guarantees
applicationtransportnetworkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-7
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-8
Multiplexingdemultiplexing
application
transport
network
link
physical
P1 application
transport
network
link
physical
application
transport
network
link
physical
P2P3 P4P1
host 1 host 2 host 3
= process= socket
delivering received segmentsto correct socket
Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)
Multiplexing at send host
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-3
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-4
Transport services and protocols provide logical communication
between app processesrunning on different hosts
transport protocols run in endsystems send side breaks app
messages into segmentspasses to network layer
rcv side reassemblessegments into messagespasses to app layer
more than one transportprotocol available to apps Internet TCP and UDP
applicationtransportnetworkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-5
Transport vs network layer
network layer logicalcommunicationbetween hosts
transport layer logicalcommunicationbetween processes relies on enhances
network layer services
Household analogy12 kids sending letters
to 12 kids processes = kids app messages = letters
in envelopes hosts = houses transport protocol =
Ann and Bill network-layer protocol
= postal service
Transport Layer 3-6
Internet transport-layer protocols
reliable in-orderdelivery (TCP) congestion control flow control connection setup
unreliable unordereddelivery UDP no-frills extension of
ldquobest-effortrdquo IP services not available
delay guarantees bandwidth guarantees
applicationtransportnetworkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-7
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-8
Multiplexingdemultiplexing
application
transport
network
link
physical
P1 application
transport
network
link
physical
application
transport
network
link
physical
P2P3 P4P1
host 1 host 2 host 3
= process= socket
delivering received segmentsto correct socket
Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)
Multiplexing at send host
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-4
Transport services and protocols provide logical communication
between app processesrunning on different hosts
transport protocols run in endsystems send side breaks app
messages into segmentspasses to network layer
rcv side reassemblessegments into messagespasses to app layer
more than one transportprotocol available to apps Internet TCP and UDP
applicationtransportnetworkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-5
Transport vs network layer
network layer logicalcommunicationbetween hosts
transport layer logicalcommunicationbetween processes relies on enhances
network layer services
Household analogy12 kids sending letters
to 12 kids processes = kids app messages = letters
in envelopes hosts = houses transport protocol =
Ann and Bill network-layer protocol
= postal service
Transport Layer 3-6
Internet transport-layer protocols
reliable in-orderdelivery (TCP) congestion control flow control connection setup
unreliable unordereddelivery UDP no-frills extension of
ldquobest-effortrdquo IP services not available
delay guarantees bandwidth guarantees
applicationtransportnetworkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-7
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-8
Multiplexingdemultiplexing
application
transport
network
link
physical
P1 application
transport
network
link
physical
application
transport
network
link
physical
P2P3 P4P1
host 1 host 2 host 3
= process= socket
delivering received segmentsto correct socket
Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)
Multiplexing at send host
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-5
Transport vs network layer
network layer logicalcommunicationbetween hosts
transport layer logicalcommunicationbetween processes relies on enhances
network layer services
Household analogy12 kids sending letters
to 12 kids processes = kids app messages = letters
in envelopes hosts = houses transport protocol =
Ann and Bill network-layer protocol
= postal service
Transport Layer 3-6
Internet transport-layer protocols
reliable in-orderdelivery (TCP) congestion control flow control connection setup
unreliable unordereddelivery UDP no-frills extension of
ldquobest-effortrdquo IP services not available
delay guarantees bandwidth guarantees
applicationtransportnetworkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-7
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-8
Multiplexingdemultiplexing
application
transport
network
link
physical
P1 application
transport
network
link
physical
application
transport
network
link
physical
P2P3 P4P1
host 1 host 2 host 3
= process= socket
delivering received segmentsto correct socket
Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)
Multiplexing at send host
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-6
Internet transport-layer protocols
reliable in-orderdelivery (TCP) congestion control flow control connection setup
unreliable unordereddelivery UDP no-frills extension of
ldquobest-effortrdquo IP services not available
delay guarantees bandwidth guarantees
applicationtransportnetworkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
networkdata linkphysical
applicationtransportnetworkdata linkphysical
logical end-end transport
Transport Layer 3-7
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-8
Multiplexingdemultiplexing
application
transport
network
link
physical
P1 application
transport
network
link
physical
application
transport
network
link
physical
P2P3 P4P1
host 1 host 2 host 3
= process= socket
delivering received segmentsto correct socket
Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)
Multiplexing at send host
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-7
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-8
Multiplexingdemultiplexing
application
transport
network
link
physical
P1 application
transport
network
link
physical
application
transport
network
link
physical
P2P3 P4P1
host 1 host 2 host 3
= process= socket
delivering received segmentsto correct socket
Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)
Multiplexing at send host
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-8
Multiplexingdemultiplexing
application
transport
network
link
physical
P1 application
transport
network
link
physical
application
transport
network
link
physical
P2P3 P4P1
host 1 host 2 host 3
= process= socket
delivering received segmentsto correct socket
Demultiplexing at rcv hostgathering data from multiplesockets enveloping data with header (later used for demultiplexing)
Multiplexing at send host
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-9
How demultiplexing works host receives IP datagrams
each datagram has sourceIP address destination IPaddress
each datagram carries 1transport-layer segment
each segment has sourcedestination port number
host uses IP addresses amp portnumbers to direct segment toappropriate socket
source port dest port
32 bits
applicationdata
(message)
other header fields
TCPUDP segment format
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-10
Connectionless demultiplexing
Create sockets with portnumbers
DatagramSocket mySocket1 = newDatagramSocket(12534)
DatagramSocket mySocket2 = newDatagramSocket(12535)
UDP socket identified bytwo-tuple
(dest IP address dest port number)
When host receives UDPsegment checks destination port
number in segment directs UDP segment to
socket with that portnumber
IP datagrams withdifferent source IPaddresses andor sourceport numbers directedto same socket
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-11
Connectionless demux (cont)
DatagramSocket serverSocket = new DatagramSocket(6428)
ClientIPB
P2
client IP A
P1P1P3
serverIP C
SP 6428DP 9157
SP 9157DP 6428
SP 6428DP 5775
SP 5775DP 6428
SP provides ldquoreturn addressrdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-12
Connection-oriented demux
TCP socket identifiedby 4-tuple source IP address source port number dest IP address dest port number
receiving host uses allfour values to directsegment to appropriatesocket
Server host may supportmany simultaneous TCPsockets each socket identified by
its own 4-tuple Web servers have
different sockets foreach connecting client non-persistent HTTP will
have different socket foreach request
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-13
Connection-oriented demux(cont)
ClientIPB
P1
client IP A
P1P2P4
serverIP C
SP 9157DP 80
SP 9157DP 80
P5 P6 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-14
Connection-oriented demuxThreaded Web Server
ClientIPB
P1
client IP A
P1P2
serverIP C
SP 9157DP 80
SP 9157DP 80
P4 P3
D-IPCS-IP AD-IPC
S-IP B
SP 5775DP 80
D-IPCS-IP B
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-15
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-16
UDP User Datagram Protocol [RFC 768]
ldquono frillsrdquo ldquobare bonesrdquoInternet transportprotocol
ldquobest effortrdquo service UDPsegments may be lost delivered out of order
to app connectionless
no handshaking betweenUDP sender receiver
each UDP segmenthandled independentlyof others
Why is there a UDP no connection
establishment (which canadd delay)
simple no connection stateat sender receiver
small segment header no congestion control UDP
can blast away as fast asdesired
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-17
UDP more often used for streaming
multimedia apps loss tolerant rate sensitive
other UDP uses DNS SNMP
reliable transfer over UDPadd reliability atapplication layer application-specific
error recovery
source port dest port
32 bits
Applicationdata
(message)
UDP segment format
length checksumLength in
bytes of UDPsegmentincluding
header
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-18
UDP checksum
Sender treat segment contents as
sequence of 16-bitintegers
checksum addition (1rsquoscomplement sum) ofsegment contents
sender puts checksumvalue into UDP checksumfield
Receiver compute checksum of
received segment check if computed checksum
equals checksum field value NO - error detected YES - no error detected
But maybe errorsnonetheless More laterhellip
Goal detect ldquoerrorsrdquo (eg flipped bits) intransmitted segment
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-19
Internet Checksum Example Note
When adding numbers a carryout from themost significant bit needs to be added to theresult
Example add two 16-bit integers
1 1 1 1 0 0 1 1 0 0 1 1 0 0 1 1 01 1 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 1
1 1 0 1 1 1 0 1 1 1 0 1 1 1 1 0 01 0 1 0 0 0 1 0 0 0 1 0 0 0 0 1 1
wraparound
sumchecksum
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-20
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-21
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-22
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-23
Principles of Reliable data transfer important in app transport link layers top-10 list of important networking topics
characteristics of unreliable channel will determinecomplexity of reliable data transfer protocol (rdt)
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-24
Reliable data transfer getting started
sendside
receiveside
rdt_send() called from above(eg by app) Passed data to
deliver to receiver upper layer
udt_send() called by rdtto transfer packet over
unreliable channel to receiver
rdt_rcv() called when packetarrives on rcv-side of channel
deliver_data() called byrdt to deliver data to upper
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-25
Reliable data transfer getting startedWersquoll incrementally develop sender receiver sides of
reliable data transfer protocol (rdt) consider only unidirectional data transfer
but control info will flow on both directions use finite state machines (FSM) to specify
sender receiver
state1
state2
event causing state transitionactions taken on state transition
state when in thisldquostaterdquo next state
uniquely determinedby next event
eventactions
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-26
Rdt10 reliable transfer over a reliable channel
underlying channel perfectly reliable no bit errors no loss of packets
separate FSMs for sender receiver sender sends data into underlying channel receiver read data from underlying channel
Wait forcall fromabove packet = make_pkt(data)
udt_send(packet)
rdt_send(data)extract (packetdata)deliver_data(data)
Wait forcall from
below
rdt_rcv(packet)
sender receiver
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-27
Rdt20 channel with bit errors
underlying channel may flip bits in packet checksum to detect bit errors
the question how to recover from errors acknowledgements (ACKs) receiver explicitly tells sender
that pkt received OK negative acknowledgements (NAKs) receiver explicitly
tells sender that pkt had errors sender retransmits pkt on receipt of NAK
new mechanisms in rdt20 (beyond rdt10) error detection receiver feedback control msgs (ACKNAK) rcvr-gtsender
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-28
rdt20 FSM specification
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
belowsender
receiverrdt_send(data)
Λ
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-29
rdt20 operation with no errors
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-30
rdt20 error scenario
Wait forcall fromabove
snkpkt = make_pkt(data checksum)udt_send(sndpkt)
extract(rcvpktdata)deliver_data(data)udt_send(ACK)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
rdt_rcv(rcvpkt) ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp isNAK(rcvpkt)
udt_send(NAK)
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Wait forACK or
NAK
Wait forcall from
below
rdt_send(data)
Λ
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-31
rdt20 has a fatal flaw
What happens ifACKNAK corrupted
sender doesnrsquot know whathappened at receiver
canrsquot just retransmitpossible duplicate
Handling duplicates sender retransmits current
pkt if ACKNAK garbled sender adds sequence
number to each pkt receiver discards (doesnrsquot
deliver up) duplicate pkt
Sender sends one packet then waits for receiver response
stop and wait
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-32
rdt21 sender handles garbled ACKNAKs
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
Wait forACK orNAK 0 udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isNAK(rcvpkt) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt)
Wait for call 1 from
above
Wait forACK orNAK 1
ΛΛ
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-33
rdt21 receiver handles garbled ACKNAKs
Wait for0 frombelow
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq0(rcvpkt)
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
Wait for1 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq0(rcvpkt)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp not corrupt(rcvpkt) ampamp has_seq1(rcvpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt)
sndpkt = make_pkt(ACK chksum)udt_send(sndpkt)
sndpkt = make_pkt(NAK chksum)udt_send(sndpkt)
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-34
rdt21 discussion
Sender seq added to pkt two seq rsquos (01) will
suffice Why must check if received
ACKNAK corrupted twice as many states
state must ldquorememberrdquowhether ldquocurrentrdquo pkthas 0 or 1 seq
Receiver must check if received
packet is duplicate state indicates whether
0 or 1 is expected pktseq
note receiver can notknow if its lastACKNAK received OKat sender
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-35
rdt22 a NAK-free protocol
same functionality as rdt21 using ACKs only instead of NAK receiver sends ACK for last pkt
received OK receiver must explicitly include seq of pkt being ACKed
duplicate ACK at sender results in same action asNAK retransmit current pkt
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-36
rdt22 sender receiver fragments
Wait forcall 0 from
above
sndpkt = make_pkt(0 data checksum)udt_send(sndpkt)
rdt_send(data)
udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) || isACK(rcvpkt1) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
Wait forACK
0sender FSM
fragment
Wait for0 frombelow
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt) ampamp has_seq1(rcvpkt)extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(ACK1 chksum)udt_send(sndpkt)
rdt_rcv(rcvpkt) ampamp (corrupt(rcvpkt) || has_seq1(rcvpkt))
udt_send(sndpkt)receiver FSM
fragment
Λ
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-37
rdt30 channels with errors and loss
New assumptionunderlying channel canalso lose packets (dataor ACKs) checksum seq ACKs
retransmissions will beof help but not enough
Approach sender waitsldquoreasonablerdquo amount oftime for ACK
retransmits if no ACKreceived in this time
if pkt (or ACK) just delayed(not lost) retransmission will be
duplicate but use of seqrsquos already handles this
receiver must specify seq of pkt being ACKed
requires countdown timer
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-38
rdt30 sendersndpkt = make_pkt(0 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
Waitfor
ACK0
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt1) )
Wait forcall 1 from
above
sndpkt = make_pkt(1 data checksum)udt_send(sndpkt)start_timer
rdt_send(data)
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt0)
rdt_rcv(rcvpkt) ampamp( corrupt(rcvpkt) ||isACK(rcvpkt0) )
rdt_rcv(rcvpkt)ampamp notcorrupt(rcvpkt)ampamp isACK(rcvpkt1)
stop_timerstop_timer
udt_send(sndpkt)start_timer
timeout
udt_send(sndpkt)start_timer
timeout
rdt_rcv(rcvpkt)
Wait forcall 0from
above
Waitfor
ACK1
Λrdt_rcv(rcvpkt)
ΛΛ
Λ
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-39
rdt30 in action
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-40
rdt30 in action
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-41
Performance of rdt30
rdt30 works but performance stinks ex 1 Gbps link 15 ms prop delay 8000 bit packet
U sender utilization ndash fraction of time sender busy sending
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
1KB pkt every 30 msec -gt 33kBsec thruput over 1 Gbps link network protocol limits use of physical resources
dsmicrosecon8bps10
bits80009
===R
Ldtrans
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-42
rdt30 stop-and-wait operation
first packet bit transmitted t = 0
sender receiver
RTT
last packet bit transmitted t = L R
first packet bit arriveslast packet bit arrives sendACK
ACK arrives send nextpacket t = RTT + L R
U sender
= 008
30008 = 000027
microsec
onds
L R
RTT + L R =
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-43
Pipelined protocolsPipelining sender allows multiple ldquoin-flightrdquo yet-to-
be-acknowledged pkts range of sequence numbers must be increased buffering at sender andor receiver
Two generic forms of pipelined protocols go-Back-Nselective repeat
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-44
Pipelining increased utilization
first packet bit transmitted t = 0
sender receiver
RTT
last bit transmitted t = L R
first packet bit arriveslast packet bit arrives send ACK
ACK arrives send nextpacket t = RTT + L R
last bit of 2nd packet arrives send ACKlast bit of 3rd packet arrives send ACK
U sender
= 024
30008 = 00008
microsecon
ds
3 L R
RTT + L R =
Increase utilizationby a factor of 3
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-45
Pipelining Protocols
Go-back-N overview sender up to N
unACKed pkts inpipeline
receiver only sendscumulative ACKs doesnrsquot ACK pkt if
therersquos a gap sender has timer for
oldest unACKed pkt if timer expires
retransmit all unACKedpackets
Selective Repeat overview sender up to N unACKed
packets in pipeline receiver ACKs individual
pkts sender maintains timer
for each unACKed pkt if timer expires retransmit
only unACKed packet
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-46
Go-Back-NSender k-bit seq in pkt header ldquowindowrdquo of up to N consecutive unACKed pkts allowed
ACK(n) ACKs all pkts up to including seq n - ldquocumulative ACKrdquo may receive duplicate ACKs (see receiver)
timer for each in-flight pkt timeout(n) retransmit pkt n and all higher seq pkts in window
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-47
GBN sender extended FSM
Wait start_timerudt_send(sndpkt[base])udt_send(sndpkt[base+1])hellipudt_send(sndpkt[nextseqnum-1])
timeout
rdt_send(data)
if (nextseqnum lt base+N) sndpkt[nextseqnum] = make_pkt(nextseqnumdatachksum) udt_send(sndpkt[nextseqnum]) if (base == nextseqnum) start_timer nextseqnum++ else refuse_data(data)
base = getacknum(rcvpkt)+1If (base == nextseqnum) stop_timer else start_timer
rdt_rcv(rcvpkt) ampamp notcorrupt(rcvpkt)
base=1nextseqnum=1
rdt_rcv(rcvpkt) ampamp corrupt(rcvpkt)
Λ
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-48
GBN receiver extended FSM
ACK-only always send ACK for correctly-received pktwith highest in-order seq may generate duplicate ACKs need only remember expectedseqnum
out-of-order pkt discard (donrsquot buffer) -gt no receiver buffering Re-ACK pkt with highest in-order seq
Wait
udt_send(sndpkt)default
rdt_rcv(rcvpkt) ampamp notcurrupt(rcvpkt) ampamp hasseqnum(rcvpktexpectedseqnum)
extract(rcvpktdata)deliver_data(data)sndpkt = make_pkt(expectedseqnumACKchksum)udt_send(sndpkt)expectedseqnum++
expectedseqnum=1sndpkt = make_pkt(expectedseqnumACKchksum)
Λ
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-49
GBN inaction
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-50
Selective Repeat
receiver individually acknowledges all correctlyreceived pkts buffers pkts as needed for eventual in-order delivery
to upper layer sender only resends pkts for which ACK not
received sender timer for each unACKed pkt
sender window N consecutive seq rsquos again limits seq s of sent unACKed pkts
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-51
Selective repeat sender receiver windows
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-52
Selective repeat
data from above if next available seq in
window send pkttimeout(n) resend pkt n restart timerACK(n) in [sendbasesendbase+N]
mark pkt n as received if n smallest unACKed pkt
advance window base tonext unACKed seq
senderpkt n in [rcvbase rcvbase+N-1]
send ACK(n) out-of-order buffer in-order deliver (also
deliver buffered in-orderpkts) advance window tonext not-yet-received pkt
pkt n in [rcvbase-Nrcvbase-1]
ACK(n)otherwise ignore
receiver
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-53
Selective repeat in action
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-54
Selective repeat dilemmaExample seq rsquos 0 1 2 3 window size=3
receiver sees nodifference in twoscenarios
incorrectly passesduplicate data as newin (a)
Q what relationshipbetween seq sizeand window size
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-55
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-56
TCP Overview RFCs 793 1122 1323 2018 2581
full duplex data bi-directional data flow
in same connection MSS maximum segment
size connection-oriented
handshaking (exchangeof control msgs) initrsquossender receiver statebefore data exchange
flow controlled sender will not
overwhelm receiver
point-to-point one sender one receiver
reliable in-order bytesteam no ldquomessage boundariesrdquo
pipelined TCP congestion and flow
control set window size send amp receive buffers
socket
door
TCP
send buffer
TCP
receive buffer
socket
door
segment
application
writes data
application
reads data
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-57
TCP segment structure
source port dest port
32 bits
applicationdata
(variable length)
sequence numberacknowledgement number
Receive windowUrg data pointerchecksum
FSRPAUheadlen
notused
Options (variable length)
URG urgent data (generally not used)
ACK ACK valid
PSH push data now(generally not used)
RST SYN FINconnection estab(setup teardown
commands)
bytes rcvr willingto accept
countingby bytes of data(not segments)
Internetchecksum
(as in UDP)
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-58
TCP seq rsquos and ACKsSeq rsquos
byte streamldquonumberrdquo of firstbyte in segmentrsquosdata
ACKs seq of next byte
expected fromother side
cumulative ACKQ how receiver handles
out-of-order segments A TCP spec doesnrsquot
say - up toimplementer
Host A Host B
Seq=42 ACK=79 data = lsquoCrsquo
Seq=79 ACK=43 data = lsquoCrsquo
Seq=43 ACK=80
Usertypes
lsquoCrsquo
host ACKsreceipt
of echoedlsquoCrsquo
host ACKsreceipt oflsquoCrsquo echoes
back lsquoCrsquo
timesimple telnet scenario
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-59
TCP Round Trip Time and Timeout
Q how to set TCPtimeout value
longer than RTT but RTT varies
too short prematuretimeout unnecessary
retransmissions too long slow reaction
to segment loss
Q how to estimate RTT SampleRTT measured time from
segment transmission until ACKreceipt ignore retransmissions
SampleRTT will vary wantestimated RTT ldquosmootherrdquo average several recent
measurements not justcurrent SampleRTT
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-60
TCP Round Trip Time and Timeout
EstimatedRTT = (1- α)EstimatedRTT + αSampleRTT
Exponential weighted moving average influence of past sample decreases exponentially fast typical value α = 0125
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-61
Example RTT estimationRTT gaiacsumassedu to fantasiaeurecomfr
100
150
200
250
300
350
1 8 15 22 29 36 43 50 57 64 71 78 85 92 99 106
time (seconnds)
RT
T
(m
illise
co
nd
s)
SampleRTT Estimated RTT
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-62
TCP Round Trip Time and Timeout
Setting the timeout EstimtedRTT plus ldquosafety marginrdquo
large variation in EstimatedRTT -gt larger safety margin first estimate of how much SampleRTT deviates from
EstimatedRTT
TimeoutInterval = EstimatedRTT + 4DevRTT
DevRTT = (1-β)DevRTT + β|SampleRTT-EstimatedRTT|
(typically β = 025)
Then set timeout interval
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-63
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-64
TCP reliable data transfer
TCP creates rdtservice on top of IPrsquosunreliable service
pipelined segments cumulative ACKs TCP uses single
retransmission timer
retransmissions aretriggered by timeout events duplicate ACKs
initially considersimplified TCP sender ignore duplicate ACKs ignore flow control
congestion control
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-65
TCP sender eventsdata rcvd from app create segment with
seq seq is byte-stream
number of first databyte in segment
start timer if notalready running (thinkof timer as for oldestunACKed segment)
expiration intervalTimeOutInterval
timeout retransmit segment
that caused timeout restart timer ACK rcvd if acknowledges
previously unACKedsegments update what is known to
be ACKed start timer if there are
outstanding segments
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-66
TCPsender(simplified)
NextSeqNum = InitialSeqNum SendBase = InitialSeqNum
loop (forever) switch(event)
event data received from application above create TCP segment with sequence number NextSeqNum if (timer currently not running) start timer pass segment to IP NextSeqNum = NextSeqNum + length(data)
event timer timeout retransmit not-yet-acknowledged segment with smallest sequence number start timer
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer
end of loop forever
Commentbull SendBase-1 last cumulatively ACKed byteExamplebull SendBase-1 = 71y= 73 so the rcvrwants 73+ y gt SendBase sothat new data is ACKed
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-67
TCP retransmission scenariosHost A
Seq=100 20 bytes data
ACK=100
timepremature timeout
Host B
Seq=92 8 bytes data
ACK=120
Seq=92 8 bytes data
Seq=
92 t
imeo
ut
ACK=120
Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
lost ACK scenario
Host B
X
Seq=92 8 bytes data
ACK=100
time
Seq=
92 t
imeo
utSendBase
= 100
SendBase= 120
SendBase= 120
Sendbase= 100
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-68
TCP retransmission scenarios (more)Host A
Seq=92 8 bytes data
ACK=100
loss
tim
eout
Cumulative ACK scenario
Host B
X
Seq=100 20 bytes data
ACK=120
time
SendBase= 120
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-69
TCP ACK generation [RFC 1122 RFC 2581]
Event at Receiver
Arrival of in-order segment withexpected seq All data up toexpected seq already ACKed
Arrival of in-order segment withexpected seq One other segment has ACK pending
Arrival of out-of-order segmenthigher-than-expect seq Gap detected
Arrival of segment that partially or completely fills gap
TCP Receiver action
Delayed ACK Wait up to 500msfor next segment If no next segmentsend ACK
Immediately send single cumulative ACK ACKing both in-order segments
Immediately send duplicate ACK indicating seq of next expected byte
Immediate send ACK provided thatsegment starts at lower end of gap
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-70
Fast Retransmit
time-out period oftenrelatively long long delay before
resending lost packet detect lost segments
via duplicate ACKs sender often sends
many segments back-to-back
if segment is lost therewill likely be manyduplicate ACKs for thatsegment
If sender receives 3ACKs for same data itassumes that segmentafter ACKed data waslost fast retransmit resend
segment before timerexpires
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-71
Host A
tim
eout
Host B
time
X
resend seq X2
seq x1seq x2seq x3seq x4seq x5
ACK x1
ACK x1ACK x1ACK x1
tripleduplicate
ACKs
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-72
event ACK received with ACK field value of y if (y gt SendBase) SendBase = y if (there are currently not-yet-acknowledged segments) start timer else increment count of dup ACKs received for y if (count of dup ACKs received for y = 3) resend segment with sequence number y
Fast retransmit algorithm
a duplicate ACK for already ACKed segment
fast retransmit
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-73
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-74
TCP Flow Control
receive side of TCPconnection has areceive buffer
speed-matchingservice matchingsend rate to receivingapplicationrsquos drain rate
app process may beslow at reading frombuffer
sender wonrsquot overflowreceiverrsquos buffer by
transmitting toomuch
too fast
flow control
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-75
TCP Flow control how it works
(suppose TCP receiverdiscards out-of-ordersegments)
unused buffer space= rwnd= RcvBuffer-[LastByteRcvd -
LastByteRead]
receiver advertisesunused buffer space byincluding rwnd value insegment header
sender limits ofunACKed bytes to rwnd guarantees receiverrsquos
buffer doesnrsquot overflow
IPdatagrams
TCP data(in buffer)
(currently)unused buffer
space
applicationprocess
rwndRcvBuffer
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-76
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-77
TCP Connection ManagementRecall TCP sender receiver
establish ldquoconnectionrdquobefore exchanging datasegments
initialize TCP variables seq s buffers flow control
info (eg RcvWindow) client connection initiator Socket clientSocket = new
Socket(hostnameportnumber)
server contacted by client Socket connectionSocket =
welcomeSocketaccept()
Three way handshakeStep 1 client host sends TCP
SYN segment to server specifies initial seq no data
Step 2 server host receivesSYN replies with SYNACKsegment server allocates buffers specifies server initial seq
Step 3 client receives SYNACK
replies with ACK segmentwhich may contain data
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-78
TCP Connection Management (cont)
Closing a connectionclient closes socket
clientSocketclose()
Step 1 client end systemsends TCP FIN controlsegment to server
Step 2 server receivesFIN replies with ACKCloses connection sendsFIN
client
FIN
server
ACK
ACK
FIN
close
close
closed
tim
ed w
ait
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-79
TCP Connection Management (cont)
Step 3 client receives FINreplies with ACK
Enters ldquotimed waitrdquo -will respond with ACKto received FINs
Step 4 server receivesACK Connection closed
Note with smallmodification can handlesimultaneous FINs
client
FIN
server
ACK
ACK
FIN
closing
closing
closed
tim
ed w
ait
closed
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-80
TCP Connection Management (cont)
TCP clientlifecycle
TCP serverlifecycle
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-81
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-82
Principles of Congestion Control
Congestion informally ldquotoo many sources sending too much
data too fast for network to handlerdquo different from flow control manifestations
lost packets (buffer overflow at routers) long delays (queueing in router buffers)
a top-10 problem
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-83
Causescosts of congestion scenario 1
two senders tworeceivers
one routerinfinite buffers
no retransmission
large delayswhen congested
maximumachievablethroughput
unlimited sharedoutput link buffers
Host Aλin original data
Host B
λout
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-84
Causescosts of congestion scenario 2
one router finite buffers sender retransmission of lost packet
finite shared outputlink buffers
Host A λin originaldata
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-85
Causescosts of congestion scenario 2 always (goodput) ldquoperfectrdquo retransmission only when loss
retransmission of delayed (not lost) packet makes larger(than perfect case) for same
λin
λout=
λin
λoutgtλ
inλ
out
ldquocostsrdquo of congestion more work (retrans) for given ldquogoodputrdquo unneeded retransmissions link carries multiple copies of pkt
R2
R2λin
λ out
b
R2
R2λin
λ out
a
R2
R2λin
λ out
c
R4
R3
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-86
Causescosts of congestion scenario 3 four senders multihop paths timeoutretransmit
λin
Q what happens asand increase λ
in
finite shared outputlink buffers
Host Aλin original data
Host B
λout
λin original data plusretransmitted data
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-87
Causescosts of congestion scenario 3
another ldquocostrdquo of congestion when packet dropped any ldquoupstream transmission
capacity used for that packet was wasted
HostA
HostB
λout
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-88
Approaches towards congestion control
end-end congestioncontrol
no explicit feedback fromnetwork
congestion inferred fromend-system observed lossdelay
approach taken by TCP
network-assistedcongestion control
routers provide feedbackto end systems single bit indicating
congestion (SNADECbit TCPIP ECNATM)
explicit rate sendershould send at
two broad approaches towards congestion control
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-89
Case study ATM ABR congestion control
ABR available bit rate ldquoelastic servicerdquo if senderrsquos path
ldquounderloadedrdquo sender should use
available bandwidth if senderrsquos path
congested sender throttled to
minimum guaranteedrate
RM (resource management)cells
sent by sender interspersedwith data cells
bits in RM cell set by switches(ldquonetwork-assistedrdquo) NI bit no increase in rate
(mild congestion) CI bit congestion
indication RM cells returned to sender by
receiver with bits intact
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-90
Case study ATM ABR congestion control
two-byte ER (explicit rate) field in RM cell congested switch may lower ER value in cell senderrsquo send rate thus maximum supportable rate on path
EFCI bit in data cells set to 1 in congested switch if data cell preceding RM cell has EFCI set sender sets CI
bit in returned RM cell
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-91
Chapter 3 outline
31 Transport-layerservices
32 Multiplexing anddemultiplexing
33 Connectionlesstransport UDP
34 Principles ofreliable data transfer
35 Connection-orientedtransport TCP segment structure reliable data transfer flow control connection management
36 Principles ofcongestion control
37 TCP congestioncontrol
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-92
TCP congestion control Goal TCP sender should transmit as fast as possible
but without congesting network Q how to find rate just below congestion level
Decentralized solution each TCP sender sets its ownrate based on implicit feedback ACK segment received (a good thing) network not
congested so increase sending rate lost segment assume loss due to congested
network so decrease sending rate
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-93
TCP congestion control bandwidth probing
ldquoprobing for bandwidthrdquo increase transmission rateon receipt of ACK until eventually loss occurs thendecrease transmission rate continue to increase on ACK decrease on loss (since available
bandwidth is changing depending on other connections innetwork)
ACKs being received so increase rate
X
X
XX
X loss so decrease rate
send
ing
rate
time
Q how fast to increasedecrease
TCPrsquosldquosawtoothrdquobehavior
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-94
TCP Congestion Control details
Sender limits rate by limiting unACKed bytesin pipeline to cwnd (like N in Go-Back-N orSelective Repeat) ie
cwnd differs from rwnd (how why) sender limited by min(cwndrwnd)
Sending rate or throughput roughly
Sender controls cwnd dynamically as afunction of perceived network congestion
rate = cwnd
RTT bytessec
LastByteSent-LastByteAcked le cwnd
cwndbytes
RTT
ACK(s)
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-95
TCP Congestion Control more details
segment loss eventreducing cwnd
timeout no responsefrom receiver cut cwnd to 1
3 duplicate ACKs atleast some segmentsgetting through (recallfast retransmit) cut cwnd in half less
aggressively than ontimeout
ACK received increasecwnd
Two phases slowstart
increase exponentiallyfast (despite name) atconnection start orfollowing timeout
congestion avoidance increase linearly
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-96
TCP Slow Start when connection begins cwnd =
1 MSS example MSS = 500 bytes
amp RTT = 200 msec initial rate = 20 kbps
available bandwidth may be gtgtMSSRTT desirable to quickly ramp up
to respectable rate increase rate exponentially
until first loss event or whenthreshold reached double cwnd every RTT done by incrementing cwnd
by 1 for every ACK received
Host A
one segment
RTT
Host B
time
two segments
four segments
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-97
Transitioning intoout of slowstartssthresh cwnd threshold maintained by TCP on loss event set ssthresh to cwnd2 (half of TCP rate when
congestion last occurred) when cwnd gt= ssthresh transition from slowstart to congestion
avoidance phase
slow start timeout
ssthresh = cwnd2cwnd = 1 MSS
dupACKcount = 0retransmit missing segment timeout
ssthresh = cwnd2 cwnd = 1 MSS
dupACKcount = 0retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKdupACKcount++duplicate ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0 congestion
avoidance
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-98
TCP congestion avoidance
Goal Approach possiblecongestion slower than inslowstart
When cwnd gt ssthreshgrow cwnd linearly ie increasecwnd by 1 MSS per RTT implementation cwnd =cwnd + MSScwnd foreach ACK received
ACKs increase cwndby 1 MSS per RTTadditive increase
loss cut cwnd in half(non-timeout-detectedloss ) multiplicativedecrease
AIMD
AIMD Additive IncreaseMultiplicative Decrease
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-99
TCP congestion control FSM overview
slow start
congestionavoidance
fastrecovery
cwnd gt ssthresh
losstimeout
losstimeout
new ACK loss3dupACK
loss3dupACK
losstimeout
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-100
TCP congestion control FSM details
slow start
congestionavoidance
fastrecovery
timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment timeoutssthresh = cwnd2
cwnd = 1 MSSdupACKcount = 0
retransmit missing segment
Λcwnd gt ssthresh
cwnd = cwnd+MSSdupACKcount = 0transmit new segment(s)as allowed
new ACKcwnd = cwnd + MSS (MSScwnd)
dupACKcount = 0transmit new segment(s)as allowed
new ACK
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3retransmit missing segment
dupACKcount == 3
dupACKcount++duplicate ACK
ssthresh= cwnd2cwnd = ssthresh + 3
retransmit missing segment
dupACKcount == 3
timeoutssthresh = cwnd2cwnd = 1 dupACKcount = 0retransmit missing segment
cwnd = cwnd + MSStransmit new segment(s) as allowed
duplicate ACK
cwnd = ssthreshdupACKcount = 0
New ACK
Λ
cwnd = 1 MSSssthresh = 64 KBdupACKcount = 0
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-101
Popular ldquoflavorsrdquo of TCP
ssthresh
ssthresh
TCPTahoe
TCPReno
Transmissionround
cwnd windowsize(insegments)
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-102
Summary TCP Congestion Control
when cwnd lt ssthresh sender in slow-startphase window grows exponentially
when cwnd gt= ssthresh sender is incongestion-avoidance phase window grows linearly
when triple duplicate ACK occurs ssthresh setto cwnd2 cwnd set to ~ ssthresh
when timeout occurs ssthresh set to cwnd2cwnd set to 1 MSS
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-103
TCP throughput
Q What is average throughout of TCP asfunction of average loss rate L RTT andbottleneck capacity C of path
For a given cwnd at any point in time
But what is average cwnd as function of L
cwnd =122 MSS
L
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
throughput =mincwnd
RTTrwnd
RTTC
$
amp
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-104
TCP throughput (more)
throughput =min122 MSS
RTT Lrwnd
RTTC
$ amp
(
Loss rate and RTTlimit throughput dueto AIMD (as cwnd
does not grow quicklyenough before halving
upon loss)
Receive window limitsthroughput due to
flow control (so as tonot overwhelm
receiver)
Bottleneckcapacity obviouslylimits throughput
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-105
Need better future TCP for ldquolong fat pipesrdquo
Motivating example 1500 byte segments 100msRTT want 10 Gbps throughput Q Will TCP achieve 10Gbps A Depends on loss rate rwnd and bottleneck capacity
To saturate capacity loss rate formula implies
Implies L lt 2 10-10 Wow Real loss rates rarely this low =gt need better TCP for
long high speed links
10Gbps lt122 MSS
RTT L=122 15008
01 L
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-106
Fairness goal if K TCP sessions share samebottleneck link of bandwidth R each should haveaverage rate of RK
TCP connection 1
bottleneckrouter
capacity R
TCP connection 2
TCP Fairness
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-107
Why is TCP fairTwo competing sessions Additive increase gives slope of 1 as throughout increases multiplicative decrease decreases throughput proportionally
R
R
equal bandwidth share
Connection 1 throughput
Conn
ecti
on 2
thr
ough
put
congestion avoidance additive increaseloss decrease window by factor of 2
congestion avoidance additive increaseloss decrease window by factor of 2
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-108
Fairness (more)Fairness and UDP multimedia apps often
do not use TCP do not want rate
throttled by congestioncontrol
instead use UDP pump audiovideo at
constant rate toleratepacket loss
Fairness and parallel TCPconnections
nothing prevents app fromopening parallelconnections between 2hosts
web browsers do this example link of rate R
supporting 9 connections new app asks for 1 TCP gets
rate R10 new app asks for 11 TCPs
gets R2
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo
Transport Layer 3-109
Chapter 3 Summary principles behind transport
layer services multiplexing
demultiplexing reliable data transfer flow control congestion control
instantiation andimplementation in theInternet UDP TCP
Next leaving the network
ldquoedgerdquo (applicationtransport layers)
into the networkldquocorerdquo