1

CIS-532 Lecture 1

Summer of 2004

2

Communication Network

• Collection of computers that are both autonomous and interconnected

• Differs from distributed computer system – in distributed system, presence of multiple

processors is made transparent to user whereas in network knowledge of the multiple processors is fundamental to user

3

Digital Versus Analog

• Classification applies to data, signaling, and transmission

• Data is “information content”

• Examples of analog data– acoustic wave amplitude as function of time for

voice– brightness of each pixel as function of time for

video

4

Digital Data

• Examples of digital data– binary data– sequence of alphanumeric characters (e-mail)

5

Signaling

• Electromagnetic waves used to carry data through channel

• Analog signaling varies continuously with time and takes continuum of values

• Digital signaling has fixed waveforms that represent the bits 0 and 1

6

Analog Signaling

• Analog signaling can be used for analog data– by using same waveform as data– by modulation

• Analog signaling can be used for digital data by modulation (modem)

7

Digital Signaling

• Can be used for digital data– by direct encoding (e.g., NRZ code)– by more complicated encoding

• Can be used for analog data– sampling– quantization

8

Analog Transmission

• Analog transmission means amplifiers used– Amplification of noise is cumulative– Only used for analog signals– Some degradation of signal must be acceptable

• e.g., local subscriber loop for telephone

9

Digital Transmission

• Repeaters used instead of amplifiers– bits are determined and signal regenerated at

each repeater– effect of noise eliminated at each stage unless

bit inversions– always used when signaling is digital– can be used for analog signaling provided

analog signal encodes digital data

10

Digitization• Nyquist Sampling Theorem

– Analog signal is uniquely determined by its samples taken at twice its maximum frequency

• this “direction” is relevant for digital encoding of analog data

• If 2^^N quantization levels, need N bits per sample

• Quantization error can be regarded as noise– SNR due to quantization is approx. 6 X N dB where N

is number of bits used to represent each sample

11

Example: Digitized Voice

• Voice grade telephone channel transmits frequencies up to 4 kHz– To convert to digital signal, sampling at 8 kHz

is needed– For telephone, SNR of 48 dB is needed.

• So 8 bits per sample are required

• Generates digital bit stream at 64 kbps (PCM channel)

12

Analog signaling for digital data– The values of an analog signal required to have

maximum frequency W may be arbitrarily specified every 1/(2W) sec-- 2W samples/sec can be specified

• this “direction” relates to analog signaling for digital data• data rate depends on number of bits per sample

– In QAM, signal constellation in plane determines number of bits encoded per sample

– Shannon’s Theorem gives upper bound on capacity (depends on SNR)

13

Evolution of Telephone Networks

• Circuit switching as opposed to dedicated circuits– Switching introduces economy of scale since

traffic for many source/destination pairs can be routed over high-capacity trunks

– Switching originally by operators, then automated mechanical, now electronic

14

Telephone Network Evol. (cont)

• Common channel signaling (CCS)– data network used by switches to exchange

control information– Separates call control from transfer of voice– Together with programmable switches, permits

value-added services (e.g., call waiting, call forwarding)

15

Telephone Network Evol. (cont)• Since 1980’s, transmission changing to SONET

(Synchronous Optical Network)– Basic STS-1 signal has rate of 51.840 Mbps

• ISDN: digital subscriber loops and service integration.– Basic access is 2B + D

• B channel is full-duplex 64 kbps. Suitable for circuit-switched connection, connection to packet-switched network, or permanent digital connection

• D channel is 16 kbps packet-switched

16

Broadband ISDN

• Integration of voice, video, data in high speed network– ATM running over SONET

17

Computer (data) Networks

• Organization of data in packets– requires control bits (headers and trailers)

• Packet switching (store and forward)– allows link bandwidth shared on as-needed

basis– superior to FDM and TDM for bursty traffic

18

Internet Protocol Hierarchy

• IP and (TCP/IP) can run over many physical networks (e.g., Ethernet, Token Ring, ATM)– allows interconnection of heterogeneous

networks– presents uniform interface to applications

• allows development of applications that are independent of physical network

19

Multiple Access Techniques

• Common channel shared by multiple stations

• Medium access control (MAC) required to transform shared channel into virtual intermittent point-to-point link

• LAN standards include Ethernet, Token Bus, Token Ring

20

Multiaccess (cont)

• MAN standards– Fiber distributed data interface (FDDI)

• similar to token ring but faster (100 Mbps)

• has timed-token mechanism that can transmit real-time traffic (voice or video) with guaranteed delay

– Distributed Queue Dual Bus (DQDB)

21

Cable Television Networks

• Current CATV uses FDM – 69 analog TV channels, each 4.5 MHz wide– Transmission over coaxial cable arranged as

unidirectional tree

• Fiber to curb (with cable to individual subscriber) can increase BW and decrease attenuation

22

CATV continued

• Migration to digital transmission is occurring to increase number of channels

• Future developments include creating LANs for subscribers to use to send reverse traffic (e.g., requests for movies)

23

Service Integration• Example: Asynchronous transfer mode (ATM)

networks– Runs over a physical layer such as SONET– 53-byte cells transmitted over virtual circuits– Different connections can be allocated different

amounts of resources (bandwidth and buffers)– User and network negotiate contract that specifies

user traffic characteristics and network guaranteed QoS

– Accomodates both real-time and nonreal-time traffic

24

Economic Issues

• Economies of scale– Due to fixed network management costs as well

as fact that cost increases slower than linearly with data rate, cost per user decreases as number of users increases

• Network externalities– Value of network service to user increases as

number of users increases

25

Economics cont.

• Due to economies of scale and network externalities, there is “critical number” of users– below critical number, subsidy is required

• Service integration– combining services in single network (e.g.,

BISDN) reduces cost of each service due to shared infrastructure

26

Network Structure

• Point-to-point versus broadcast– in point-to-point, each link connects pair of

nodes– in broadcast, all nodes share common link

(channel). All users receive each packet sent but only those for whom it is addressed retain it

– more generally, network may support multicasting

27

Local Area Networks (LANs)

• Typically use broadcast link(s)– require multiaccess algorithms

• Maximum distance between hosts is limited– implies upper bound for propagation delay--

which is important to multiaccess algorithms

• Examples: Ethernet, Token ring, Token bus

• MANs are similar to LANs but larger area

28

Wide Area Networks (WANs)

• End systems (hosts) are interconnected via communication subnet

• Subnet consists of switching nodes (routers) connected by transmission lines (links)

• May be packet switched or circuit switched

29

Wireless Networks

• Radio, packet radio, microwave, satellite

• May or may not involve host mobility and time-varying network topology– Example: cellular radio systems– Example: wireless LANs

30

Hybrid Networks

• Both wireless and wireline components– Example: satellite-fiber networks– Example: wired LAN on aircraft with flying

router having wireless connection to terrestrial network

31

Messages

• Message is a single unit of communication in sense that it is useful to recipient only if completely delivered– Example: file in file transfer system– Example: image in image transfer system– Example: one line of symbols in interactive

terminal session

32

Messages, cont.

• Concept of messages not very useful for voice or video– flow model corresponding to stream of bits is

more appropriate– stream may be CBR or VBR

33

Packets

• Messages are broken up into units of manageable size called packets– packets are transmitted as strings of bits together

with additional control bits– control bits may indicate addresses, offsets, etc.– if packets have constant length, then called cells

• Before being broken into packets, messages may be transformed for purpose of data compression and/or encryption

34

Connectionless Versus Connection-Oriented Service

• Services provided by a layer may be connectionless or connection-oriented– for transport layer refers to messages– for network layer refers to packet

35

Connectionless Service

• Messages (or packets) are independent of each other– analogous to postal service– order of messages (packets) need not be

preserved– generally not reliable, but may be made reliable

through use of acknowledgements • analogous to certified return-receipt postal service

36

Connection-Oriented Service

• Messages (packets) are part of a connection set up (and later terminated) between communicating hosts– Messages are delivered in order– Service is generally reliable: no duplication or

omission of messages– Analogous to telephone service

37

Characteristics of Traffic

• Traffic arrival rate and variability

• Connection duration

• Distribution of message length

• Allowable delay and variability of delay

• Required reliability

38

Examples of Traffic Types

• Interactive terminal-to-computer sessions– low message rate– message length short– delay requirement moderately strict– required reliability high

39

Traffic Examples, cont.

• File transfer sessions– message rate low– message length very long– delay requirement very relaxed– required reliability very high

40

Traffic Examples, cont.

• Packetized voice– concept of message not applicable– bit arrival rate moderate– delay requirement stringent (especially jitter)– required reliability low

41

Circuit Switching

• When session is set up, path is chosen and bandwidth allocated on each link (by FDM or TDM).– If no path with sufficient BW, call is rejected– Advantage: once call is accepted, BW is

guaranteed; no queuing– Disadvantage: inefficient utilization of

transmission capacity if traffic is bursty

42

Packet Switching

• Store and forward

• Statistical multiplexing– No fixed allocation of BW– Packets from different sessions combined into

single queue for each outgoing link– Full transmission capacity of link dedicated to

single packet

• Advantage: full utilization of link capacity whenever traffic is present

43

Connectionless versus Connection-Oriented Routing

• Virtual circuit routing– connection-oriented– fixed path (but not fixed BW) assigned at start

of session; all packets follow same path– Example: ATM

• Datagram routing– packets in session are routed independently– Example: IP