chapter 6: the transport layer

what is the purpose of the transport layer

to deliver data across networks to processes with the desired reliability or quality, irrespective of lower layer technologies used for each hop

what is a transport entity

the software and/or hardware that does the bulk of the work

what are the two types of transport service

1. connection-oriented
2. connectionless

why aren’t the transport and network layer merged into one

the transport layer exists entirely on users’ machines but the network layer runs on routers operated by carriers

the existence of the transport layer allows its service to be more reliable than the underlying network

what is another name for the TPDU

segments

which layers are the transport service providers

layers 1-3

which layers are the transport service users

layers 4 + 5

what are the 5 primitives for a connection-oriented transport interface

1. LISTEN
2. CONNECT
3. SEND
4. RECIEVE
5. DISCONNECT

true or false: both data link and transport protocols deal with:

* error control
* sequencing
* flow control

true

true or false: in point-to-point links, the transport does not need to explicitly address the destination

false

what does TSAP stand for

Transport

Service

Access

Point

what are TSAPs

specific endpoints in the transport layer

true or false: IP addresses are an example of a TSAP

false: they are an example of an NSAP

true or false: application processes can attach themselves to a local TSAP to establish a connection to a remote TSAP

true

what is the purpose of a portmapper

user connects to the portmapper, which listens to a well-known TSAP, user then sends a message specifying the service name and the port mapper returns the TSAP address and the user releases the connection with the portmapper to establish one with the desired service

what happens in the portmapper when a new service is created

the new service must register itself with the portmapper giving both its:

* service name (usually ASCII string)
* TSAP

so portmapper can record this info in its internal database to be used later

what is the intial connection protocol

each machine that wishes to offer services to remote users has a special process server to act as a proxy for less heavily used servers

a set of ports are simultaneously listened to, waiting for a connection request

how do throwaway transport addresses work

each time a transport address is needed, a new one is generated

after connection release, the address is discarded and never used again

why are throwaway transport addresses be used

delayed duplicate packets never find their way to a transport process

what is one problem throwaway transport addresses introduce

harder to connect with a process in the first place

what are three ways packet lifetime can be restricted to a known maximum

1. restricted network design
2. putting a hop counter in each packet
3. time-stamping each packet

how does restricted network design restrict packet lifetime

packets are prevented from looping, combined with some way of bounding delay including congestion over the longest possible path

how does putting a hop counter in each packet restrict packet lifetime

count down from some value each time packet forwarded across a hop until counter reaches 0 and packet discarded

how does time-stamping each packet restrict packet lifetime

each packet shows the time it was created

routers agree to discard any packet that is time-stamped older than some agreed upon time

what are 3 things the transport protocol deals with

1. error control
2. sequencing
3. flow control

similar to link layer protocols

true or false: both buffering and flow control are required in the transport and data link layer

true

what is the name of the protocol used during connection establishment

the three-way handshake

how does the three way handshake work

1. CONNECTION REQUEST: host 1 sends a sequence number (x) as starting point for a counter to uniquely identify segments from host 1 to 2
2. CONNECTION ACCEPT: host 2 sends an ACK of seq no. (x) from H1 and also a seq number (y) used to count segments from H2 to H1
3. CONNECTION: H1 sends ACK for seq no. (y) from H2 and starts sending data as part of the connection with its seq. no. (x)

true or false: each new connection uses a different pair of sequence numbers

true: allows us to sport duplicate segments and delayed segments belonging to an old connection between the same 2 end points

what are the two types of connection release

1. asymmetric
2. symmetric

how does asymmetric connection release work

one of the hosts terminates the connection without notice to the other end point

what are the downsides of asymmetric connection release

one-sided abrupt connection termination could lead to loss of data if received after DR has been sent

how does symmetric connection release work

1. H1 sends a DR request to H2 and starts timer
2. H2 sends any remaining data then sends a DR and starts timer
3. H1 receives H2 DR, is certain H2 has no more to send and sends the final ACK so H2 can also release connection

what is a benefit of symmetric connection release

less probability of data loss

what is a problem with symmetric connection release

the two-army problem

how are error cases handled in symmetric connection release

through ACKs, timers and retransmissions

true or false: in symmetric connection release initiated by H1, H2 will never retransmit a DR

true: only releases connection after a timeout because it knows H1 has no more data to send

what is UDP

User Datagram Protocol is a ‘best effort’ connectionless service used without ACKs by the Internet in the transport layer

how big is a UDP header

8 bytes

what fields are in a UDP header

1. source port address
2. destination port address
3. UDP length
4. UDP checksum

true or false: UDP checksum is computed for the header, the data and the conceptual IP pseudo header

true

what fields are contained in the IPv4 psuedoheader for both UDP and TCP

1. source IP address
2. destination IP address
3. protocol number
4. UDP length

true or false: UDP deals with flow control, congestion control and retransmission upon receipt of a bad segment

false

when is UDP especially useful

client-server situations or for applications that require precise control over:

* packet flow
* error control
* timing

what is TCP

Transmission Control Protocol is a connection based protocol designed to provide a reliable byte stream between exactly 2 processes over an unreliable internetwork

what does TCP guarantee for a working network

delivery of ALL data (without any losses), without errors and in the correct order

true or false: it is required for TCP to send data fast enough to use capacity without causing congestion

true

what is a socket

end points connecting end points between sender and receiver using TCP service

what does each socket consist of

* host IP address
* 16-bit number local to host (port)

true or false: a socket may only be used for one connection at the same time

false

what is a well-known port

ports reserved for standard services that can only be started by privileged users

what is the range of numbers for well-known ports

ports with numbers below 1024

true or false: TCP does supports multicasting or broadcasting

true

true or false: all TCP connections are full duplex and point-to-point

true

is a TCP connection:

* a byte stream
* a message stream

a byte stream: message boundaries are not preserved end-to-end

true or false: every byte on a TCP connection has its own 32-bit sequence number

true

how many bytes is the fixed part of a TCP header:

20-bytes

what limits the size of a TCP segment

* each segment must fit in 65,515-byte IP payload
* the Maximum Transfer Unit (MTU) which each segment must fit into

what is path MTU discovery

ICMP error messages are used to find the smallest MTU for any link on the path in a VC in TCP so an upper bound on segment size can be set to avoid fragmentation

why is path MTU discovery used

some network paths may have small MTUs which may require a IP packets carrying TCP segments to be fragmented, which degrades performance

finding smaller MTU on a path allows this to be avoided

what basic protocol is used by TCP entities

sliding window protocol with a dynamic window size

what is included in the source and destination port in the TCP header

* TCP port
* host’s IP address

48-bit unique end point

what do the source and destination end points identify together in the TCP header

together they identify the connection

what information does the connection identifier contain and in what format is it in TCP

it is a 5 tuple containing:


1. the protocol (TCP)
2. source IP
3. source port
4. destination IP
5. destination port

what information does the sequence number field in the TCP header hold

counts bytes NOT segments, i.e. if segment seq no = 100 and it is 512 B long, next segment seq no. will be 612 to indicate position of payload’s 1st byte in stream

what information does the acknowledgement number field in the TCP header hold

indicates next in-order byte expected in a cumulative acknowledgement system because it summarises the received data with a single number

how long are the acknowledgement and sequence number fields in a TCP header

32 bits

true or false: TCP uses piggy-backed ACKs

true

what information does the TCP header length field contain

how many 32-bit words are contained in the header

what field is used to indicate that an ACK has been sent in the TCP header

ACK field: set to 1 when segment carries ACK, set to 0 otherwise

what is the purpose of the RST field in the TCP header

abruptly terminates connection due to an unrecoverable error (maybe due to host crash)

also rejects an invalid segment or refuses a connection request

what is the purpose of the SYN field in the TCP header

used to establish connections:

* CONNECTION REQUEST: SYN = 1 , ACK = 0
* CONNECTION ACCEPTED: SYN =1 , ACK = 1

what is the purpose of the FIN field in the TCP header

used to release connections: asserted to show when sender has no more data to transmit

what is the purpose of the window size field in the TCP header

FLOW CONTROL: specifies how many bytes may be sent, starting at the byte acknowledged

what does a window size field = 0 in the TCP header mean

bytes up to and including ACK no. -1 have been received but receiver has not had enough time to process, pauses data stream

what options may be used in the TCP header

* Max Segment Size (MSS)
* timestamp : carried by sender, echoed by receiver
* window scale : sender and receiver negotiate a window scale factor at start of connection
* Selective ACK (SACK) : allows receiver to send range of seq no.s that have been received (often used w congestion control