TTM4100 Komm., Tjen. og Nett Chapter 1 - Computer networks and the internet

1.1 What is the internet?

A Nuts.and-Bolts description

- Host (end system), server, mobile, router, link-layer switch, modem, base station, smartphone, cell phone tower.

- End systems are connected together by a network of communication links and packet switches. Communication links can be cable, copper wire, optical fiber and radio.

- Packages of data are sent trough these communication links. - A packet switch takes a packet arriving at on of its communication links and sends it onward in

one of its outgoing communication links. The two most common packet switches are routers and link-layer switches.

- Just like trucks moving from a warehouse to its delivery location using roads and intersections, data travels in packages trough a computer network.

- Internet Service Providers (ISPs) is what end systems access the internet trough. - End systems, packet switches and other pieces of the internet runs protocols that control the

sending and receiving of information on the internet. Transmission Control Protocol (TCP) and internet Protocol (IP) are two of the most important. IP specifies the format of the packages.

- To make sure that everyone agrees on what each protocol means the Internet standards was developed. They define the protocols.

A Service description

- Internet is an infrastructure that provides services to applications. Electronic mail, web surfing, social networks, instant messages, Voice-over-IP (VoIP), video

streaming, distributed games, peer-to-peer (P2P) file sharing, tv, remote login etc. - Application Programming Interface (API)

API is a set of rules that the sender needs to follow so that the Internet can deliver data to the destination, much like the postal service has rules for how to send a letter.

What is a protocol?

- A protocol defines a set of actions to take when sending and receiving messages, much like how people communicate, saying 'Hi' first, and based on the response decides what to do next.

1.2 The network edge - Things connected at the edge of the internet is called end systems (computers, etc). They are

also referred to as hosts because they host applications. Hosts are divided into two categories, servers and clients.

- Access network is the network connecting edge systems to the first router, known as edge router.

- Frequency-division multiplexing (FDM), splitting up the frequency spectre so that several things can use the same wire (DSL using the same as the telephone).

- Hybrid fiber coax (HFC), connecting trough the cable television companys existing cable. Uses a special modem, to exchange between analog signals and digital.

- Fiber, promises much faster speed than cable and DSL.

- Twisted-pair copper wires is the most used wires, it is two wires twisted around each other. This is done to prevent interference from other wires. The width and length of the wire decides the bandwidth, reaching up to 10 Gbps.

- Fiber optics are better on long distance, it uses flashes of light to represent bits. It has a higher cost.

- LTE (Long -Term Evolution).

1.3 The network core - Most packet switches apply store-and-forward transmission, meaning they must receive the

whole packet before it can begin transmitting it. - Each packet switch has multiple links attached to it, and each link has its own output buffer. If an

arriving packet finds a link busy with another packet, its placed in the output buffer to wait for the link to open up. If the output buffer is already full, either the incoming packet or the packet in the buffer willl be dropped, this is called packet loss.

- Each router has a forwarding table, mapping destinations to outgoing links. The internet has a number of special routing protocols, using some shortest path algorithm to

decide the shortest path from each router to every destination.

- There are two fundamental approaches to moving data through a network of links and switches: Circuit switching and packet switching. In circuit, the resources needed along a path (buffers, link transmission rate) to provide

communication are reserved, in packet s they are not. Telephone is an example of circuit switching.

- The internet works as a network of networks. It is build on several layers of ISPs (internet service providers). No ISP is present in every city in the world, rather, the top layer consist of about 10-15 Tier-1 ISPs, which provide to region ISPs and so forth. An ISP can multi-home into an ISP provider above it self, meaning it connect its router to more than one of the ISP providers routers. This means that it can continue transmitting data even if one of the providers has a failure.

- Internet Exchange Point (IXP), a meeting point where ISPs can peer together.

1.4 Delay, Loss and Throughput in Packet-Switched Networks - A packet encounters several types of delay at each node: nodal processing delay, queuing

delay, transmission delay and propagation delay. All adding up to a total nodal delay. - Processing delay, the time to examine the packets header and determine where to direct it.

Microseconds or less. - Queuing delay, the time while a packet waits to be transmitted onto the link.

In practice it varies between micro- to milliseconds. - Transmission delay, a packet can only be transmitted after all before-coming packets have

been transmitted. The delay is the bit-length of packet/transmission rate between router A and B. (L/R) micro- to milliseconds.

- Propagation delay, the time required to propagate from A to B. Depends on the medium which it travels trough. The delay is the distance/propagation speed.

1.5 Protocol Layers and Their Service Models - Protocols are put together in layers, one layer performs certain actions within that layer and uses

the services of the layer directly below it. - A layer n protocol can be distributed among the end system, packet switches and other

components of the network. - One drawback of this layering model is that one layer might duplicate lower-level functionality. - Five-layer internet protocol: Physical - Link - Network - Transport - Application

Application layer, here lies HTTP, SMTP and FTP. A packet of information at this level is called message.

Transport layer, TCP and UDP. A transport-layer packet is called a segment. Network layer, moves network-layer packets, known as datagrams, from one host to another.

Includes the IP Protocol. All network layers must run the IP Protocol. This layer is often referred to as the IP-layer.

Link layer, this layer routes a datagram trough all the routers from sender to receiver. At each node the network layer passes the datagram down to the link layer who which then delivers it to the next node along the route. A packet at this level is called a frame.

Physical layer, moves the individual bits from one node to another. - When taken together, the protocols of the various layers are called the protocol stack, like the

internet protocol stack above.

1.6 Networks Under Attack - Viruses are malware that require some form of user interaction to infect the users device, for

example an email attachement. - Worms are malware that can enter a device without explicit interaction. - A packet sniffer stores a copy of every packet that flies by it.

Because packet sniffers are passive they are hard to detect. Use encryption, because it is likely that someone is reading your packets.

1.7 History of computer networking and the internet - This part tells the story of how the internet evolved from the research in packet-switches in the

1960s, trough the development of TCP and IP and into the internet explosion in the 1990s when the world wide web first popped up.

Chapter 2 - Application layer

2.1 Principles of Network Applications - At the core of network applications is writing programs that run on different end systems and

communicate with each other over the network. - When designing a network application one must choose an application architecture to use.

Normally one would use either Client-server architecture or peer-to-peer architecture. - We always label one of the communicating parts as a client and the other as host. On a p2p

architecture, the host that is downloading is labeled as client and the one uploading is labeled as server.

- We talk about client and server processes and the communication between two processes operating on divergent end systems. A process sends and receives messages trough a software interface called a socket. A process can be seen as a house and the socket as its door.

- A socket is the interface between the application layer and the transport layer within a host. It is also referred to as the Application Programming Interface (API)

- The internet provides more than one transport-layer protocol, so you have to choose which one to use. You choice is based on what your application needs, we can broadly classify the possible services along four dimensions: reliable data transfer

- If a protocol guarantees that sent data is delivered correctly and completely to the other end of the application it is said to provide reliable data transfer.

- Some loss-tolerant applications, like multimedia, doesnt need this. throughput

- The transport-layer protocol can guarantees a certain amount of throughput. Applications that have throughput requirements are said to be bandwidth-sensitive applications (Multimedia), the ones who does not are called elastic applications (mail, file transfer).

timing - Guarantees that bits pumped into the socket arrives at the receiving end in no more than,

for example, 100 msec later. This is good for real-time applications (Skype). security

- Encryption and decryption.

- The internet provides two types of transport-layer protocols, TCP () and UDP (User datagram protocol).

- TCP is a connection-oriented service and reliable data-transfer service. Connection-oriented means that there is a handshake between the client and server before

data begins to flow. After a handshake a TCP connection is said to exist between the two. Messages can be sent both ways at the same time during a connection.

The communication process can rely on CP to deliver all data sent with our error and in the proper order.

- UDP is a no-frills, lightweight transport protocol, providing minimal services. UDP is connectionless, so there is no handshaking. It does not provide reliable data transfer, so data may be lost and may arrive out of order.

- Todays internet cannot provide any timing or throughput guarantees, but there are services that provide satisfactory service for applications in need of such guarantees. More in chapter 7.

- An application-layer protocol defines: The type of messages exchanged. The syntax of the various message types. The semantic of the fields (the meaning of the fields) Rules for determining when and how a process sends messages and responds to messages.

- Some application-layer protocols are specified in RFCs (a document stating internet standards) and therefore in the public domain, for example HTTP. If a browser follows the HTTP RFC then it will be able to retrieve web pages from any server also following the rules of HTTP RFC.

- It is important to distinguish between network applications and application-layer protocols. An application-layer protocol is only a piece of a network application.

An example is the Web consisting of many components: a standard for document format (HTML), web browsers, web servers, and an application-layer protocol (HTTP).

2.2 The Web and HTTP - HTTP is implemented in two programs, a client program and a server program. The client

program and server program, operating on different end systems, talk to each other by exchanging HTTP messages. HTTP defines the structure of these messages and how they are exchanged.

- HTTP uses TCP. - The socket interface is the door between the client and the server. The client sends HTTP

request to the socket interface and receives HTTP responses back. - The server does not store any state information about the client, so if a client requests the same

file twice in a row the server will respond with the same file twice instead of saying 'I just sent you that file'. HTTP is said to be stateless protocol

- Using TCP the developer must decide on either sending all request/responds on one TCP connection or use separate TCP connections for each request. E.g persistent connection or non-persistent connection.

- HTTP reuest message:

- The first line is called the request line, the subsequent lines are called the header lines. The request line has three fields: method, URL and HTTP version.

- HTTP response message

- It has three sections: a status line, six header lines and then the entity body. The status line has three fields: protocol version, status code and corresponding status message.

- HTTP uses cookies, this makes it possible for a website to keep track of users. There is a cookie header line in both the request and response, on the users end system there is a file keeping track of the cookies and backend there is a database also keeping track.

- A web cache, also called a proxy server, is a network entity that satisfies HTTP requests on the behalf of an origin Web server. The web cache has its own disk storage and keeps copies of recently requested objects in this storage.

- A cache is both a server and a client at the same time. A client request is sent to the proxy server, if it does not have the requested object in memory it opens a TCP connection with the server and retrieves it from there, it then stores the object in cache and retrieves it to the client. If it has it in memory it retrieves it right away.

- Usually a cache is bought and installed by an ISP or for example by a university.

- A web cache reduces the response time on a client requests, it can also reduce the traffic on an institutions link to the internet.

- A problem with the use of cache is that the copy might be out of date. This is solved by using a conditional get, which let the cache verify that it has the latest version. It uses the Last-Modified field to do this.

2.3 File Transfer: FTP - Runs on top of TCP just as HTTP, but there are some differences. - FTP uses two parallel TCP connections, a control connection and a data connection. The

control connection remains open through the whole session but a new data connection is established for every file.

- FTP must keeps track of the clients state, therefore the number of connections is limited compared to HTTP which is stateless.

2.4 Electronic Mail in the Internet - Email uses the SMTP (simple mail transfer protocol). - SMTP does not normally use an intermediate mail server, even if the senders server is on the

opposite side of the world compared to the receiver, a TCP connection is made directly between the two. If the receivers server is down, the message is kept in the senders mail server queue until it can be sent.

- A mail access protocol (POP3, IMAP) is used to transfer mail from the recipients mail server to the recipients user agent.

- IMAP is used to create a remote folder structure which can be accessed with more than one device. It has commands that permit a user agent to obtain only parts of messages, this is good if there is a low bandwidth and the user only wants to download say the head of each mail.

- When using web mail the user communicates with the mail server using HTTP.

2.5 DNS-The Internets Directory Service - Internets hosts can be identified in many ways.

Hostname: cnn.com, www.yahoo.com etc. IP address: 127.7.54.34 (fours bytes)

- DNS(domain name system) translates between hostname and IP addresses. DNS is a distributed database implemented in a hierarchy of DNS servers, and an application-layer protocol that allows hosts to query the database. DNS runs over UDP and uses port 53.

- DNS adds a delay to the internet application using it, but usually the desired IP address is often cached in a 'nearby' DNS server.

- DNS is also used for Host aliasing, Mail server aliasing and Load distribution. The latter meaning that the server retrieving IP address of a busy site (distributed over more than one server) rearranges the order so that user are distributed over all the IP addresses of that site.

- DNS also uses caching, since IP addresses often are changed the cache is refreshed after a certain amount of time (usually two days). The time is decided based on the type.

- Read more on page 164-170.

2.6 Peer-to-Peer Application - In P2P file distribution, each peer can redistribute any portion of the file it has received to any

other peer, thereby assisting the server in the distribution process. - The distribution time is the time it takes to get a copy of the file to all N peers.

Read about formula on p. 172.

- DHTs (distributed hash tables), each peer will only hold a small subset of the totality of the key, value pairs.

2.7 Socket Programming: Creating Network Applications - There are two types of network applications.

One type is an implementation whose oprerations is specified in a protocol, it is said to be 'open' because all rules specifying its operations are well known. If one developer programs the server program and one programs the client program, they will work together if they both carefully follows the rules of the RFC.

The other type is a proprietary network application. Client and server program employ an application-layer protocol that has not been published in any RFC.

- You have to choose between UDP or TCP. - Using UDP a destination address must be attached to the packet before it can be pushed out the

door. When a socket is created a port number is attached to it, this number as well as the IP address of the receiver is what is used in the destination address. Example with code on page 186.

- Using TCP we cannot just send packets out the door, TCP i a connection-oriented protocol so there needs to be a handshake. With the TCP connection established one side just drops the packet it wants to send into the TCP connection via its socket (Not like UDP where a destination needed to be attached). With the server running the client can initiate a TCP connection by creating a TCP socket specifying the server address. When the socket is established the client initiate a three-way handshake and establishes a TCP connection with the server. The three-way handshake takes plays in the transport-layer and is invisible to the client and server programs.

Chapter 3 - Transport Layer

3.1 Introduction and Transport-Layer Services - A transport-layer protocol provides for logical communication between application processes

running on different hosts. - The transport layer turns the application-layer messages into transport-layer packet, known as

transport-layer segments. - The transport-layer moves messages from application processes to the network edge (network-

layer), it doesnt have anything to do with how the messages are moved within the network core. Consider the example with the two families sending letters to each other, each family has one person who is responsible for collecting and sending (via the postal service) outgoing letters and distributing incoming letters to the right persons. This person is like the transport-layer protocol and the postal service is like the network-layer protocol.

- The possible services a transport-layer protocol can provide are often constraint by the network-layer protocol. If the postal service cannot guarantee a certain maximum delivery time, nor can the responsible person in each household.

- IP (Internet Protocol) is a 'best-effort delivery system', it makes no guarantee that the segments will get delivered. It is an unreliable service.

- Extending host-to-host delivery to process-to-process delivery is called transport-layer multiplexing or demultiplexing.

- TCP provides reliable data transfer using flow control, sequence numbers, acknowledgments, and timers. TCP converts IPs unreliable service between end systems to reliable service between processes.

- TCP also provides congestion control, meaning links and routers in the connection wont get swamped with excessive amount of traffic. This is done by regulating how much the sending side of the connection can into the network.

3.2 Multiplexing and Demultiplexing - The transport-layer is responsible for delivering the incoming segments from the network-layer to

the right socket, it does this by examining a set of fields which each transport-layer segment has. This job of delivering transport-layer segments to the correct socket is called demultiplexing.

- The job of gathering data chunks at the source host from different sockets, encapsulating eachh data chunk with header information to create segments, and passing them to the network-layer is called multiplexing.

- Transport-layer multiplexing requires that sockets have unique identifiers and that each segment have special fields that indicates the socket which it is to be delivered. These fields are named the source port number field and the destination port number field.

- Using UDP this process is straight forward. The client wraps the message in a destination address containing IP address and port number to the recipient, the network layer encapsulate the segment in an IP datagram and makes a best-effort to deliver it. If it reaches its destination, the transport layer of that host reads the destination port number and forwards it to the socket identified by that port number.

- The difference with TCP is that it uses a four-tuple (source IP address, source port number, destination IP address, destination port number) when demultiplexing an incoming segment.

3.3 Connectionless Transport: UDP - A problem with the lack of congestion control is that if many people are streaming media content

via UDP, many links and routers will have so much overflow that almost no data goes through. It will also cause the TCP senders to dramatically reduce their sending rates.

- It is possible for an application to have reliable data transfer with UDP, that is if the reliability is built into the application (for example with acknowledgement/retransmission mechanisms).

3.4 Principles of Reliable Data Transfer - Reliable data transfer protocols based on positive acknowledges and negative acknowledges

(data is good or data is not good) are known as ARQ (Automatic Repeat reQuest) protocols. ARQ also has three additional protocol capabilities to handle the presence of bit errors:

- Error detection - Receiver feedback.

Examples are the positive (ACK) and negative (NAK) acknowledgement replies. - Retransmission.

- When ACK and NAK are being used we run into another problem. What if the ACK or NAK is corrupted? This can be checked with a checksum bit, but how should the protocol recover from such an error. One approach is to just resend the last packet, but this introduces duplicate packets. This is a problem for the receiving end because it cannot know if the next packet is new data or a retransmission. The solution to duplicate packets is a sequence number field.

- If, in addition to losing bits, the underlying channel also can lose packets as well, we have more concerns to address. Namely, how to detect packet loss and how to handle it. We can handle it by using checksum, sequence number, ACK packets and retransmission,

just like above. The detection is handled by the sender. A retransmission wait-time is set, the number is based on probability and statistic, so if that time passes without the sender receiving an ACK the probability that the packet is lost is high and the packet is therefore retransmitted. This sometimes leads to duplicate packets, but we can handle that ass described earlier.

- We have now assembled the key elements of a data transfer protocol: checksum, sequence number, timers, ACK and NAK. We now have a working reliable data transfer protocol.

- This stop-and-wait system introduced is slow! If we introduce pipelining we can speed up the process. Pipelining has the following consequences for reliable data transfer protocol: The range of sequence number must be increased. The sender and receiver sides of the protocols may have to buffer more than one packet. The range of these two will depend on the manner in which a DTP responds to lost, corrupted

and overly delayed packets. The two basic approaches are Go-back-N and selective repeat.

- In a Go-Back-N (GBN) protocol, the sender has a maximum of unacknowledged packets it can be waiting for. Read more on page 247.

- The GBN protocol allows the sender to potentially fill the pipeline with packets, which is good. But GBN can in some scenarios itself suffer from performance problems. If the window size and bandwidth delay are both large, a single packet error can cause GBN to retransmit a large number of packets.

- Selective-repeat protocols avoid unnecessary retransmissions by having the sender retransmit only those packets that it suspects were lost or corrupted at the receiver.

3.5 Conneciton-Oriented Transport: TCP - A TCP connection provides a full-duplex service, meaning if there is a TCP connection

between host A and host B, data can flow freely both ways at the same time. It is also point-to-point, that is, between a single sender and a single receiver.

- When data comes through the door, TCP places it in the send-buffer. The maximum amount of data that can be grabbed from this buffer and placed in a segment is limited by the maximum segment size (MSS).

- The TCP segment consist of header fields and a data field. The data field contains a chunk of application data.

- Two of the most important fields in the TCP segment header are the sequence number field and the acknowledgment number field. The sequence number for a segment is the byte-stream number of the first byte in the segment.

- TCP only acknowledges bytes up to the first missing byte in the stream, TCP is said to provide cumulative acknowledgments (if 0-500 and 1000-1500 has been received, it still says it is waiting for 501).

- Each time TCP retransmit it sets the timeout interval to twice the previous value. When the timer is started again after a successful transmit it is again derived from recent values of estimatedRTT and DevRTT. This provides a limited form of congestion control.

- If the TCP receives three duplicate ACKs for the same data, it takes this as an indication that the data has been lost. It then performs a fast retransmit, retransmitting the missing segment before that segments timer expires.

- TCPs error-recovery is best characterised as a hybrid between GBN and SR protocols.

- Each host in a TCP connection has a receive buffer. Received bytes that are in the correct order are placed there for the application to read. The application does not necessarily read data at the moment it arrives. If the application is slow at reading the data, the sender can very easily overflow the buffer. To stop this TCP provides a flow-controls service, it makes sure that the sender cannot send to much data. This is very similar to congestion control but is obviously done for other reasons. The sender maintain a variable called receive window, it gives the sender an idea of how

much free buffer space is available at the receiver. From time to time the receiver sends the amount of free space to the sender, the sender then makes sure that the amount of unacknowledged data < free space in the receivers buffer.

The receiver only attaches the information on free space to data or ACKs it is sending to the sender, so if the buffer fills up and the sender stops sending data we have a problem. The receiver will not be able to tell when the space has freed up. To solve this problem TCP makes the sender continue to send segments with one data byte when the buffer is full.

- When a TCP connection is established a flag-bit in the header called SYN is set to 1. This indicates that a connection is being set up. on the third message (the second from client to server) and for the rest of the connection lifetime the SYN bit is set to 0.

- For the connection to be closed both Hosts must send a segment containing a FIN bit set to 1 and both must acknowledge the other hosts shutdown message.

3.6 Principles of Congestion Control - At the broadcast level, we can distinguish among congestion control approaches by whether the

network layer provides an explicit assistance to the transport layer for congestion control purposes: End-to-end congestion control Network-assisted congestion control

- For network-assisted congestion control, congestion information is fed back from the network to the sender in one of two ways. Direct feedback may be sent from a network router to the sender, this takes the form of a

choke packet, saying "im congested!". The router can also mark the segment going to the receiver, upon receipt the receiver notifies

the sender.

3.7 TCP Congestion Control - TCP uses end-to-end congestion control, since the IP layer provides no explicit feedback to the

end systems regarding network congestion. - TCP perceives the amount of congestion and adjust its send rate accordingly. This is done by

looking at the number of loss events (either a timeout or three duplicate ACKs). - TCP starts slow and increases the sending rate every time a segment is acknowledged, it does

this until a loss event occurs. - More about the three phases (slow start, congestion avoidance and fast recovery) on page 301. - Tcps congestion control consist of linear increase in cwnd (congestion window size) of 1 MSS

(maximum segment size) per RTT and then halving of cwnd on a loss event. TCP congestion control is therefore often referred to as an additive-increase, multiplicative decrease (AIMD) form of congestion control.

Chapter 4 - The network layer

4.1 Introduction

- Forwarding The process of sending a packet from input on a router to the correct output link. Every router has a forwarding table, this table is used to decide which link the data should be

forwarded to. It is the routing algorithm that determines the values that are inserted into the forwarding table.

- Routing The process of sending a packet trough a network from the sender to the receiver.

- Packet switch A device that transfers a packet from input link interface to output link interface. For ex. a link-layer switch and a router.

- In some computer networks there is a third network-layer function (the first two are forwarding and routing), connection setup. Some network-layer architectures require the routers on the path from sender to receiver to handshake before data is sent. Much like the three-way handshake. ATM, frame relay and MPLS.

- Network service model Defines the characteristics of end-to-end transport of packets between sending and receiving

end systems.

- CBR (constant bit rate) ATM (asynchronous transfer mode ) network service delay, variability in end-to-end delay (jitter) and the fraction of cells that are lost or delivered

late are guaranteed to less than a specified value. - ABR (Available bit rate) ATM network service

cells cannot be reordered, but may be lost.

4.2 Virtual Circuit and Datagram Networks intro - The network-layer provides both connectionless and connection services. Connection uses

handshaking. - Virtual Circuit Networks

Networks that only provide connection services.

- A VC consist of a path(links and routers) between source and host ,VC numbers, in number for each link along the path, and entries the forwarding table in each router along the path. A packet belonging to a VC will carry a VC number in its header. Because a virtiual circuit may have a different VC number on each link, each intervening router must replace the VC number of each traversing packet with a new VC number. The new VC number is obtained from the forwarding table.

- Three phases in a virtiual circuit: VC setup Data transfer VC teardown

- Datagram networks Networks that only provide connectionless services.

- VC and DN are two fundamental classes of computer networks. Each time an end system wants to send a package it stamps it with the address and sends it

into the network. Each router has a forwarding table that links destination addresses to interfaces. because forwarding tables can be modified any time, different packages sent from one system

to another can take different routes trough the network.

4.3 Whats inside a router? - Four router components can be identified:

Input ports. - Performs the physical layer function of terminating an incoming physical link at a router. - Also performs link-layer functions needed to interoperate with the link layer at the other side

of the incoming link. - The lookup function is also performed here (consulting the forwarding table).

Switching fabric. - Connects the input ports to the output ports.

Output ports. - Stores packets received from the switching fabric and transmits these packets on the

outgoing link by performing the necessary link-layer and physical-layer functions. Routing processor.

- Execute the routing protocols, maintains routing tables and attached link state information and commutes the forwarding table for the router.

- A routers input ports, output ports and switching fabric together implement the forwarding function and are almost always implemented in hardware. These are often collectively referred to as the router forwarding plane.

- Switching can be accomplished in a number of ways. Switching via memory. Input and output function as traditional I/O devices in a traditional

operating system. This was done in the simplest and earliest routers, but is still being used with some modifications.

switching via a bus. An input port transfers a packet directly to the output port over a shared bus, without intervention by the routing processor. This is normally done by attaching some header information with output port number. The packet is received by all ports but only the one matching the header will keep it. Only one packet can use the bus at any given time.

switching via an interconnection network. A crossbar switch is an interconnection network consisting of 2N buses that connect N inputs ports to N output ports. Each vertical bus intersects each horizontal bus at a crosspoint, which can be opened or closed at any time by the switch fabric controller.

- Output port processing takes packets that have been stored in the output ports memory and transmits them over the output link. This includes selecting and equeueing packets for transmission, and performing the needed link-layer and physical-layer transmission functions.

- A consequence of output port queuing is that a packet scheduler at the output port must choose one packet among those queued for transmission. This can be done in either a simple manner or width some weighting taken in to consideration.

4.4 The Internet Protocol - The IPv4 datagram format ncludes the

following key fields: Version number. This decides how

the router can determine the rest of the fields.

Header length. Type of service. This is used to

distinguish real-time-, high throughput-, and reliability datagrams.

Datagram length. Total length of IP datagram (header plus data)

Identifier, flag, fragmentation offset. Time-to-live (TTL). Included to

ensure that datagrams does not circulate forever. Decremented by 1 each time it is processed by a router.

Protocol. Indicates the specific transport-layer protocol to which the data portion of this IP datagram should be passed. Used only on the end of the journey. Binds the network layer and transport layer together.

Header checksum. Detects bit errors in a received IP datagram. Source and destination IP addresses. Options. Allows the IP header to be extended. Dropped from IPv6. Data (payload). Raison dtre for the datagram. Contains the transport-layer segment to be

delivered.

- The maximum amount of data that a link-layer frame can carry is called the maximum transmission unit (MTU), since IP datagrams are encapsulated inside link-layer frames the MTU places a hard limit on the length of IP datagrams.

- Because of this a datagram might suddenly be to big to be sent over a link on the route. The solution then is to fragment the datagram into two or more smaller datagrams. Each of these smaller datagrams are referred to as a fragment. The job of reassembling is given to the end systems, so that the network core is kept simple. The identifier, flag and fragmentation offset is used to determine how to piece things back together. The last fragment has the flag bit set to 0, the rest has it set to 1. If one or more of the fragments are lost the incomplete datagram is discarded and not sent to the transport layer.

- A host typically only has one link into the network. The boundary between the host and the physical link is called an interface. Because a routers job is to pass datagrams from one link to another it must necessarily have atlas two links and therefore more than one interface. IP requires each host and router interface to have its own IP address. Thus, an IP address is technically associated with an interface, rather than with the host or router containing that interface.

- IP addresses are managed under the authority of the Internet Corporation for Assigned Names and Numbers (ICANN). ICANN also manages the DNS root servers.

- Dynamic Host Configuration Protocol (DHCP) allows a host to obtain an IP address automatically. DHCP is often referred to as the plug-and-play protocol.

- The NAT (network address translation) router behaves to the outside world as a single device with a single IP address. Every home device connected to the NAT router sends data to the internet trough the same IP address. In practice the NAT-enabled router is hiding the home network from the outside world. The router know which internal host to direct incoming data to base on the NAT translation table and port numbers as well as IP addresses in the table entries. Many people object to the use of NAT because ports should be used for addressing processes, routers are supposed to process packets only up to layer 3, and the NAT protocol

violates the so-called end-to-end argument (hosts should be talking directly to each other). They also say that we should use IPv6 to solve the lack-of-addresses problem.

- NAT interfere with P2P because a TCP connection cant be set up with a host inside a NAT, because it cannot act as a server (that is, a hack called connection reversal solves this problem, UPnP).

- Internet Control Message Protocol (ICMP) is used by hosts and routers to communicate network-layer information to each other. ICMP messages have a type and a code field, and contain the header and the first 8 bytes of the IP datagram that caused the ICMP message to be generated (for example an error message when a router does not find a route to your destination).

- IPv6 has been designed to remove the problem of running out of IP addresses, as well as some other things. The changes are evident in the datagram format. Expanded addressing capabilities.

128 bits instead of 32. A streamlined 40-byte header. Flow labeling and priority. IPv6

makes it possible to differentiate between datagrams to give some priority. This might be used for real-time services or for customers paying more.

- These things have been dropped from IPv4: Fragmentation/Reassembly. IPv6 does not allow for fragmentation, if a router receives an IPv6

datagram to big it simply drops it and sends a "Packet to big" ICMP message back to the sender.

Header checksum. The designer of IPv6 felt that since TCP/UDP and link-layer protocols performs checksum there was no need for IPv6 to do it as well.

Options.

- Transitioning from IPv4 to IPv6 is hard because IPv4-capable systems are not capable of handling IPv6. But there is some options. Declare a flag-day when all devices are shut of and upgraded. This is unrealistic. Gradually integrating IPv6 hosts and routers into an IPv4 world. A dual-stack approach, where IPv6 nodes also have a complete IPv4 implementation. Such a

node can send/receive both types. The problem here is that if one node along the road is only IPv4 friendly then the datagram has to use the IPv4 header and since there are some fields in the IPv6 that have no counterpart in IPv4 they will be lost.

An alternative is tunneling, which solves the above mentioned problem. It basically puts the entire IPv6 datagram and puts it in the data field of an IPv4 datagram.

4.5 Routing Algorithms - The job of routing is to determine good paths from sender to receiver, through the networks of

routers. Looking at in like a graph with weighted edges, it all boils down to finding the shortest path. It is a little more tricky than a simple shortest path algorithm, because there are more factors playing a role. For example datagrams from a host in network A should not be sent through network B, but rather through C and so forth.

- A routing algorithm is either global or centralized. A global routing algorithm computes the least-cost path using complete, global knowledge

about the network. Referred to as link-state (LS) algorithms, since the alg. must be aware of the cost of all links in the network.

In a decentralized routing algorithm, the calculation of the least-cost path is carried out in an iterative, distributed manner. No node has complete information about the cost of all network links. Instead, each node begins with only the knowledge of the cost of its on directly attached links. Referred to as a distance-vector (DV) algorithm.

- A routing algorithm can also be classified as either static (slow changing, manually by humans) or dynamic (fast changing, runs either periodically or topological).

- It can also be either load-sensitive or load-insensitive. In a load-sensitive algorithm the link cost vary dynamically based on the amount of congestion.

- Link-state algorithm can be Dijkstras or Prims. Page 394. - Distance-vector algorithm is Bellman-Ford.

- Hot-potato routing (get rid of the datagram as fast as possible) is used so that autonomous systems (ASs, a group of routers) can find the least expensive gateway router to send its datagram.

- Many ISPs partition their network into multiple ASs.

Chapter 5 - The Link Layer: Links, Access, Networks, and LANs

5.1 Introduction to the Link Layer - Every device that runs a link-layer protocol is referred to as a node. That includes hosts, routers,

switches, and WiFi access points. Every communication channel that connects nodes is referred to as links.

- Over a given link, a transmitting node encapsulates the datagram in a link-layer frame and transmits the fram into the link.

- The details of provided services can vary from one link-layer protocol to the next. Possible services that can be offered by a link-layer protocol include: Framing. Encapsulation of the datagram. Each frame consist of a detailed and a number of

header fields. Link access. A medium access control (MAC) protocol specifies the rules by which a frame is

transmitted onto the link. This is interesting when more than two nodes share the same link. Reliable delivery. Guarantees to move the datagram over the link without errors using

acknowledgment and retransmission. This is often used in links that are prone to high error rates, such as wireless link.

Error detection and correction.

- The link layer is implemented in a network adapter, also sometimes known as a network interface card (NIC).

- Most of the link layer is implemented in hardware, but a part is implemented in software that runs on the hosts CPU. The software components of the link layer implement higher-level link-layer functionality such as assembling link-layer addressing information and activating the controller hardware.

- The link layer is the place in the protocol stack were hardware meets software.

5.2 Error-Detection and -Correction Techniques - Even with the use of error-detection bits there still may be uneducated bit errors. The methods

that has the least probability of failing often has the largest overhead (more computation is needed).

- Here are three methods of detecting errors. Parity checks. Uses a single parity bit which is set to 1 if the number of 1s in the data is even

and 0 if the number is odd. Something called two-dimensional parity is better, it thinks of the data as a matrix and stores one bit for each row and for each column. It can then not only detect if an error has occurred but can also fix it because it knows what bit. The ability of the

receiver to both detect and correct errors is known as a forward error correction (FEC). FEC reduces the number of retransmissions.

Checksumming Methods. The d-bits data are treated as a sequence of k-bit integers. They are summed and the sum is used for error detection. The Internet checksum is based on this technique. Checksumming requires little overhead (TCP/UDP checksums uses only 16 bit). This is mostly used in the transport-layer because error detection is implemented in the software and therefore needs to be executed fast, but in the link-layer it is the hardware doing the job and it can work much faster, therefore a better method is used (CRC).

Cyclic redundancy check (CRC). CRC codes operates as followed. Consider the d-bit piece of data, D, that the sending node wants to be send to the receiving node. The sender must first agree on an r + 1 bit pattern, known as a generator, which is denoted G. The key idea is that for a given amount of data, D, the sender will choose r additional bits, R, and append them to D such that the resulting d + r bit pattern is exactly divisible by G (no remainder) using modulo-2 arithmetic. When the data reaches the receiver it checks to see if this is still the case, if not it knows an error has occurred.

5.3 Multiple Access Links and Protocols - A point-to-point link consist of a single sender at one end and a single receiver on the other. Two

protocols using this is the point-to-point protocol (PPP) and high-level data link (HDLC) - A broadcast link can have multiple sending and receiving nodes all connected to the same link.

Here we have the multiple access protocols. Almost all MAPs can be classified as either channel partitioning protocols, random access protocols or taking-turns protocols.

- Channel partitioning protocols. Recall from section 1.3 that time-division multiplexing and frequency-division multiplexing are two techniques that can be used to partition a broadcast channels bandwidth among all nodes sharing that channel. TDM divides time into time frames and further divides time frames into N time slots. Slot sizes are chosen so that a single packet can be transmitted during a slot time. This is not good when a node has to wait for its turn even though it is the only node sending frames.

- FDM shares both the advantages and drawbacks of TDM. - A third option is code division multiple access (CDMA). CDMA assigns a different code to

each node, making it possible for them all to send data at the same time.

- In a random access protocols, a transmitting node always transmits at the full rate of the channel, R bits. When there is a collision, each node involved repeatedly retransmits its frame until its frame gets through without a collision. Each node waits a random amount of time before retransmitting.

- One RAP is slotted ALOHA. It devides time into slots. When a node has a frame to send it waits until the start of the next slot and sends the frame, if there is a collision it is detected before the slots end and retransmitted in the next slots with a probability of p ( between 0 and 1). A slot which exactly one node transmits is called a successful slot. The efficiency of ALOHA is based on the percentage of slots being successful slots.

- In the carrier sense multiple access (CSMA) and CSMA with collision detections (CSMA/CD) carrier sending (listen before speaking) and collision detecting (if someone else begins talking at the same time, stop talking) are both embodied.

- The efficiency of CSMA/CD is the long-run fraction of time during which frames are being transmitted on the channel without collisions when there is a large number of active nodes, with each node having a large number of frames to send.

- With taking turn protocol there is always one node transmitting R bps. There are ceveral taking-turn protocols. Polling protocol. One node is designated as a master node and polls each of the nodes in a

round-robin fashion. The master node tells node 1 that it can transmit up to a maximum number of frames, next it does the same with node 2 (the master node can determine when a node has finished by observing the lack of signals on the channel). This eliminates the collisions and empty slots delay, but has a few drawbacks: a polling delay, and if the master node fails the entire channel becomes idle.

Token-passing protocol. There is no master node. Instead a token (I can send) is passed between the nodes. A node only keeps the token if it has frames to send. Still we have the problem of a single node failing causing the entire channel to crash.

5.4 Switched Local Area Network - A link-layer address is also known as a LAN

address, a physical address or a MAC address. (MAC being the most opopular)

- Host and routers have link-layer addresses (as well as network-layer addresses). More correctly it is their adapter (network interface) that have link-layer addresses. A Host or a router with multiple network interfaces will thus have multiple link-layer addresses.

- A link-layer switch do not have MAC addresses associated with their interfaces that connect to hosts and routers. It only carries datagrams between hosts and routers transparently.

- The MAC address is 6 bytes long, giving 2^48 possible MAC addresses. They were designed to be permanent, but it is possible to change the MAC address via software.

- No two adapters have the same MAC address. This is possible because IEEE manages the MAC address space and a company manufacturing devices which needs MAC addresses buys them in chunks from IEEE.

- MAC addresses have a flat structure, meaning that it does not matter were the device goes, the MAC address stays the same (unlike IP addresses).

- When a sender wants to send a frame to some destination it inserts the MAC address into the frame and sends it into the LAN. Sometimes frames are broadcasted, meaning everybody gets it. The adapter therefore checks the MAC address of the frame when it is received to see if it should extract the data. If a sending adapter wants all other adapters in the LAN to receive the frame it inserts a special MAC broadcast address (usually 48 consecutive 1s, FF-FF-FF-FF-FF-FF in hex) into the destination address field of the frame.

- The Address Resolution Protocol (ARP) has the job of translating between MAC addresses and the network-layer addresses (for example IP addresses). An ARP module in the sending host takes any IP address on the same LAN and returns the corresponding MAC address. Each host and router has an ARP table in its memory, it contains IP address, MAC address and a time-to-live field.

- If an ARP table does not contain the IP-MAC address then the host broadcast out an ARP packet including the senders IP- and MAC address and the known IP address of the receiver, the one with the matching IP address sends back an ARP packet with the desired mapping. The senders ARP table is then updated.

- ARP is probably best considered a protocol that straddles the boundary between the link and network layers.

- Consider the case where a host wants to send a datagram to another host on another subnet. The sending host must first send the datagram to the router interface on the router connecting the to subnets, so the frame needs to include that address. It also needs to include the address of the receiving host in the other subnet, but the sender does not know the MAC address so it uses the IP address. The interface on the 'correct side of the router (which has its own ARP table) finds the matching MAC address and forwards the datagram to the correct host in the other subnet.

- Today, Ethernet is by far the most prevalent wired LAN technology. - The original ethernet LAN used a coaxial bus to interconnect the nodes. - By the 90s most companies used Ethernet. The hosts are directly connected to a hub with

twisted-pair copper wire. A hub is a physical-layer device that acts on individual bits rather than

frames. When a bit arrives from one interface, the hub simply re-creates the bit, boosts it energy strength, and transmits the bit onto all the other interfaces.

- in the 2000s the hub was replaced with a switch, which was collision-less. More on the switch later.

- The sending adapter encapsulates the IP datagram with an Ethernet frame and passes the frame to the physical layer. The receiver extracts the datagram and passes it to the network layer.

- The fields of the Ethernet frame are: Data field. Carries the IP

datagram. The maximum transmission unit (MTU) of Ethernet is 1500 bytes.

Destination address. Contains the MAC address, 6 bytes.

Source address. MAC address of the sending adapter.

Type field. The receiving adapter needs to know which network-layer protocol (it doesn't have to be IP) it should pass the content of the data field.

Cyclic redundant check. Used to detect bit errors in the frame. Preamble (8 bytes). The first 7 bytes has a value of 10101010 and the last is 10101011. The

first 7 serve to 'wake up' the receiving adapters and to synchronize their clocks. The last to bits of the 8 byte says 'important stuff is coming'.

- All of the Ethernet technologies provide connectionless service. It is also unreliable. It does retransmit only if the application uses TCP.

- The Ethernet standard includes the CSMA/CD protocol. But in a switch-based Ethernet LAN there are no collisions and, therefore, there is no need for a MAC protocol.

- Link-layer switch. The role of the switch is to receive incoming link-layer frames an forward them onto outgoing links. The switch itself is transparent to the hosts and routers in the subnet. The rate of which frames arrive to any of the switchs output interface may exceed the link capacity, to handle this the output interfaces have buffers.

- Filtering is the switch function that determines whether s from should be forwarded or dropped. Switch filtering and forwarding is done with a switch table. An entry in the switch table contains a MAC address, the switch interface that leads towards the MAC address and the time that the entry was placed in the table. Note that switches forward packets based on the MAC address rather than on IP addresses. When the switch receives a frame from interface x with a MAC address, it uses that address to lookup its table, there is three possible outcomes: There is no entry with that MAC address. In this case the switch broadcasts the frame. There is an entry in the table, associating the MAC address with interface x. Because the

MAC address belongs to a adapter in the LAN segment it came from there is no need to forward it, and it is dropped.

There is an entry in the table, associating the MAC address with interface y != x. The frae forwards the frame to interface y.

- The switch table is built automatically, dynamically and autonomously, switches can be said to be self-learning. The table is initially empty, but for each incoming frame the switch stores the MAC address from the source field, the interface it came from and the time. The switch deletes an entry if no frames are received from that address after some period of time (the aging time).

- Switches are also full-duplex, meaning they can send and receive at the same time.

- Properties of a Link-Layer Switch: Elimination of collisions. Heterogenous links. Because the switch isolates each link, they can operate at different

speeds and can run over different media. Management. A switch eases network management. For example if an adapter malfunctions

and continually sends frames the switch can detect that and disconnect the malfunctioning adapter.

- What are the pros and cons of routers and switches? First consider switches.

- pros They are plug-and-play. Can have relatively high filtering and forwarding rates. Switches only have to process up

through layer 2, whereas routers have to process datagrams up through layer 3. - cons

The active topology is restricted to a spanning tree to prevent cycling of broadcasted frames.

A large switch network would require a large ARP table, causing a substantial amount of RP traffic and processing.

Susceptible to broadcast storms - if one host goes crazy and transmits an endless stream of frames the switch would forward all of these frames, causing the entire network to collapse.

Now consider routers - pros

Packets do not normally cycle because of the hierarchical structure.

Not restricted to a spanning tree, can therefore use the best way from host to host.

Provide firewall protection against broadcast storms.

- cons Not plug-and-play. The host

that connect to them need their IP address to be configured.

Has a larger per-packet processing time than switches.

- Virtual local area networks (VLANs) solves three problems (Lack of traffic isolation, inefficient use of switches, managing users) of LAN. The VLAN can be set up and easily configured by a network manager. Say a company is using this to split up its network into different VLANs. They are then isolated and cannot send messages to each other. This problem is fixed by using a single device that contains both a VLAN switch and a router. The router can then receive from one department and forward it to another department. (read page 510 for a clearer picture)

- A more scalable approach to interconnecting VLAN switches is known as VLAN Trunking. The trunk port belongs to all VLANs, and frames sent to any VLAN are forwarded over the trunk link to other switches. The standard Ethernet frame has a four-byte VLAN tag that carries the identity of the VLAN to which the frae belongs. The tag is added into the frame by the sending switch and processed and removed by the receiving switch. .

5.5 Link Virtualization: A Network as a Link Layer - Multiprotocol label switching (MPLS) evolved to improve the forwarding speed of IP routers by

using a fixed-length label. Another advantage is that a router using MPLS knows more than one way through the network (with IP only one way is stored in the table), therefore a form of traffic engineering can be done.

- MPLS has been used to implement virtual private networks (VPNs)

5.6 Data Center Networking - In recent years companies like Google, Microsoft and more has build huge data centers. Each

center has its own data center network that interconnects its host with each other and the Internet.

- The cost of large data centers is huge. So networking innovations is the key to reducing the overall cost and maximizing performance.

- The hosts in a data center, called blades and resembling pizza boxes, are generally commodity hosts that include CPU, memory and disk storage. They are stacked in racks, each rack having typically 20-40 blades. At the top of each rack is a switch, named top of the rack (TOR) switch, that interconnects the hosts in the rack with each other and with other switches in the data center.

- For a small data center, a simple network consisting of a border router, a load banlancer and a few tens of racks all interconnected by a single ethernet switch could suffice. But to scale to tens to hundreds of thousands of hosts, a data center often employs a hierarchy of routers and switches, such as the picture on the right (there can be many more access routers).

- Some companies deploy another topology. If all Tier-2 switches are connected to all tier-1 switches then connections between internal hosts will never have to go above tier-1 switches, drastically increasing the capacity of the system.

5.7 Retrospective: A Day in the Life of a Web Page Request - Bob connects to the schools network through an Ethernet cable. This is what happens.

1. Bobs OS creates a DHCP request messaging puts it in an UDP segment with destination port 67 and source port 68. The UDP segment is then placed within an IP datagram with a broadcast IP destination and a source IP destination of 0.0.0.0 since Bob haven't received an IP address.

2. The IP datagram is placed within an Ethernet frame. The frame has a destination MAC address of ff-ff-ff-ff-ff-ff so it will be broadcasted. The source MAC address is that of Bobs laptop.

3. This frame is the first sent by Bobs laptop to the Ethernet switch. 4. The router receives the frame and extracts the datagram, the datagrams payload (UDP

segment) is demultiplexed up to UDP, and the DHCP message is distracted. The DHCP server now has the DHCP message.

5. The DHCP server allocates an IP address to Bobs laptop and creates a DHCP ACK message containing this address, as well as the IP address of the NDS server, the IP address of the default gateway router and the subnet block. The DHCP message is put inside an UDP segment, which is put inside an IP datagram, which is put inside an Ethernet frame. The frame contains two MAC addresses, the sender (router) and receiver (Bob).

6. The frame is sent (unicast) by the router to the switch. The switch has already (from receiving the frame from Bob) learned that the MAC address of Bobs laptop should be forwarded on the output port leading to him.

7. Bobs laptop receives and extracts the message, and stores the IP address of itself as well as the IP address to the DNS. It also installs the Ip address of the default gateway into its IP forwarding table. Bobs laptop will send all datagrams with destination outside of its subnet to the default gateway. The laptop is now ready to fetch Web pages.

- When Bob types www.google.com into his web browser, the laptop starts by creating a TCP socket that will be used to send the HTTP request. In order to create the socket he will have to know the IP address of google. DNS provides this.

- Read all steps on page 523-526.

Chapter 6 - Wireless and Mobile Networks

6.1 Introduction - We can identify the following elements in a wireless network:

Wireless hosts. Hosts are the end-systems devices that run applications. Wireless links. A host connects to a base station or to another wireless host through a

wireless communication link. Base station. A base station is responsible for sending and reviewing data to and from a

wireless host that is associated with that base station. Host associated with a base station is often said to be in infrastructure mode. When a mobile device moves between areas it often have to change base station, this process is called handoff.

Network infrastructure. This is the larger network with which a wireless host may wish to communicate.

- At the highest level we can classify wireless network according to two criteria: (1) whether a packet in the wireless network crosses exactly one wireless hop or multiple wireless hop. (2) whether there is infrastructure such as a base station in the network: Single-hop, infrastructure-based. Single-hop, infrastructure-less. Multi-hop, infrastructure-based Multi-hop, infrastructure-less

6.2 Wireless Links and Network Characteristics - If we replace the wired Ethernet with a wireless network, a wireless network interface would

replace the hosts wired Ethernet interface. No changes would be needed at the network layer or above.

- We can find several differences in the link layer: Decreasing signal strength. Electromagnetic radiation loses strength as i passes through

material, even just the air. Interference from other sources. Multipath propagation occurs when portion of the electromagnetic wave reflect off objects and

the ground, taking paths of different lengths between a sender and receiver. This results in blurring of the signal.

- This means that bit errors are more common in wireless networks. Therefore, wireless link protocols employ both powerful CRC error detection codes and link-level reliable-data-transfer protocols that retransmits corrupted frames.

- The signal-to-noise ratio (SNR) is a relative measure of the strength of the received signal and this noise. A larger SNR makes it easier for the receiver to extract the transmitted signal from the background noise.

- The hidden terminal problem and fading of a signals strength is when two stations, A and C, are both sending to station B. A and Cs signals are not strong enough to detect each others transmission, yet they are strong enough to interfere with each other at station B.

- Code devision multiple access (CDMA) belongs to the family of channel partitioning protocols (access protocol).

- CDMA encodes the data being transmitted so that different senders can still send on the same frequency and at the same time. The receiver has the 'key that unlocks the message.

6.3 WiFi: 802.11 Wireless LANs - The fundamental building block is the basic serve set (BSS). It contains one ore more wireless

stations and a base station, known as an access point (AP). - Each wireless station needs to associate with an AP before it can send or receive network data. - When a network administrator installs an AP, the administrator assigns a one- or two-word

Service Set ID (SSID) to the access point. - A WiFi jungle is a is any physical location were a wireless station receives a strong signal from

two ore more APs. - APs regularly sends beacon frames, including SSID and MAC address. The wireless station on

a device scans the 11 channels looking for those beacons and that is how it finds out which wireless networks are available.

- Inspired by Ethernet and its random access protocol, wireless LAN uses a random access protocol referred to as CSMA with collision avoidance (CSMA/CA). "carrier sense multiple access", meaning that each station senses the channel before transmitting, and does not transmit when the channel is busy. Due to hidden terminal problem and fading this is not straightforward but the steps below explains the process. Distributed Inter-frame Space (DIFS) is the time from the station senses the channel is idle

until it start sending a frame. If it is not idle, a random backoff value is chosen. This value is counted down while the

channel is idle, if it is busy the time is frozen. When the counter reaches zero the station transmits and wait for an acknowledgment. If an acknowledgment is received and the station has more frames to send it starts at step 2. If

it is not received the station reenters the backoff phase in step 2, with the random value chosen from a larger interval.

- Differently from CSMA/CD the CSMA/CA protocol does not start sending at the moment it senses the channel to be idle, it counts down the backoff time first. This is done because CSMA/CA cannot detect collision and if two channels start sending at the same time then they will both send the entire frame (and collide), waisting a lot of time. This can also lead to collision if the two stations are hidden rom each other or they chose backoff times that are close to each other.

- In order to avoid the problem with hidden terminals, the IEEE 802.11 protocol allows a station to use a short Request to Send (RTS) control frame and a short Clear to send (CTS) control frame to reserve access to the channel.

- This can improve performance in two ways: The hidden station problem is mitigated. Because the RTS and CTS frames are short, the collision

involving them will last only the duration of the short RTS or CTS frame.

- It also introduces delay and consumes channel resources, so the RTS/CTS is only used (if at all) to reserve the channel for the transmission of a long data frame.

- The 802.11 frame is similar to the link-layer frame but they have some differences. Address 2 is the MAC address

of the station that transmits the frame.

Address 1 is the MAC address of the wireless station that is to receive the frame.

To understand address 3 remember that the BSS is part of a subnet. Address 3 is the MAC address to the interconnecting routers interface.

Duration. The time reserved for sending the frame. The type and subtype fields are used to distinguish between RTS, CTS, ACK and data

frames. The to and from fields are used to define the meaning of the different address fields. The WEP field indicates whether encryption is being used or not.

6.4 Cellular Internet Access - Read the book!

Chapter 7 - Multimedia Networking

7.1 Multimedia Networking Application - We define a multimedia networking application as any network application that employ audio or

video.

7.2 Streaming Stored Video - Streaming video applications can be classified into three categories:

UDP streaming - With UDP streaming, the server transmits video at a rate that matches the clients video

consumption rate by clocking out the video chunks over UDP at a steady rate. - UDP uses normally a small client-side buffer holding no more than a second of video. - UDP encapsulate the data chunk within transports packet designed for transporting video

and audio, using the real time transport protocol (RTP). - The client also sends messages to the server with info regarding pauses, resume,

reposition and so on. HTTP streaming

- The video is simply stored in an HTTP server as an ordinary file with a specific URL. - A full client application buffer indirectly imposes a limit on the rate that video can be sent

from server when streaming over HTTP. - HTTP make use of the HTTP byte-range header in the HTTP GET request message, it

specifies the range of bytes the client currently wants to receive. This is useful when jumps in the video is performed.

Dynamic Adaptive HTTP streaming (DASH) - In DASH, the video is encoded into several different version, with different qualities level. - The client dynamically request chunks of video segment of a few seconds in length from the

different versions based on its current bandwidth. - Each version is stored in the HTTP server, each with a different URL. The server also has a

manifest file with its different versions and the related bandwidth . The client request this file and learns which version it should use when.

- The two latest are the most used.

- Client buffering is when the client stores a part of the video not yet shown in the buffer, to absorb variations in server-to-client delay.

7.3 Voice-over-IP - Most existing VoIP runs over UDP, including Skype (unless the user is behind a firewall blocking

UDP). - The time from when a packet is generated at the source until it is received at the receiver can

fluctuate from packet to packet. This phenomenon is called jitter. Jitter can often be removed by adding sequence number, timestamps and a playout delay.

- It is not easy to recover from packet loss and retransmitting lost packet may not be feasible in a real-time conversational application such as VoIP.

- Because of this VoIP often uses some sort of loss anticipation scheme: Forward Error Correction (FEC). The basic idea of FEC is to add redundant information to the

original packet stream. For the cost of marginally increasing the transmission rate, the redundant information can be used to reconstruct approximations or exact versions of some of the lost packets.

Interleaving. Meaning sending unit 1,5,9,13 in one chunk and 2,6,10,14 in another. This means that if one chunk is lost then it is less noticeable. This increases latency and may therefore be better for streaming stored audio.

Error concealment. The packet that came before the lost packet is duplicatet and replaces the lost packet.