.

IN
C
Faculty:

N
SY
BCS502 – COMPUTER NETWORKS Prof. Ashok Herur

U
ashok. herur @eastpoint. ac.
VT in
 Module 4

.IN
C
N
Transport Layer

SY
U
VT

Computer Networks - Prof Ashok Herur 2

 Main topics in Module 4

• The Transport service

.IN
• Different transport layer protocols – TCP and UDP

C
N
• Flow control and Error control

SY
• Congestion control

U
VT

Computer Networks - Prof Ashok Herur 3

OSI Reference model

.IN
C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 4

 Functions of the Transport Layer

• It is the lowest host-to-host protocol.

.IN
• Facilitates communication between host processes.

C
N
• Divides the message into data segments on the sending side, and reassembles

SY
them on the receiving side.

U
• Takes complete responsibility for a reliable, end-to-end transport of the
VT
message.

• Packets can be: Corrupted, or Lost, or Delivered out of sequence.

• Also takes care of Flow control between the hosts.

Computer Networks - Prof Ashok Herur 5
Segments in the Transport Layer

.IN
C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 6

 Transport entity

• The software and hardware within the Transport layer that does the work is

.IN
called the Transport entity.

C
• The Transport entity is generally located (when it comes to Internet) in the

N
SY
Operating System, or in a library package bound into network applications.

• It can also be in a separate user process or even in the Network Interface card.

U
VT

Computer Networks - Prof Ashok Herur 7

 Transport layer addressing

• When an application process wishes to set up a connection to a remote

.IN
application process (on another host), it must specify it unambiguously.

C
• In the Internet, such addresses are called Ports. (in generic terms, an address

N
SY
is called TSAP – Transport Service Access Point).

• This is similar to an IP address in the Network layer (NSAP – Network Service

U
Access Point).
VT
• However, note that a host could be running many applications simultaneously,
in which case a single IP address will lead to many Port addresses.

Computer Networks - Prof Ashok Herur 8

Transport layer addressing

.IN
C
The Transport layer at the

N
destination host should

SY
know the port to which
the data is to be delivered.

U
VT Eg. shown is from TSAP
1208 on Host 1 to TSAP
1522 on Host 2.

Computer Networks - Prof Ashok Herur 9

 Connection-oriented vs. Connectionless sessions

• The Network layer too had Connectionless (Datagram Packet Switching) and

.IN
Connection-oriented (Virtual Circuit Packet Switching) transmissions.

C
• However, they were restricted to transmission from the transmitting router

N
(representing the transmitting host) to the terminal router (representing the

SY
destination host).
• Any deficiencies in the service has to be handled by the Transport layer.

U
VT
• Transport layer has two main protocols:
• TCP (Transport Control Protocol) – Connection-oriented
• UDP (User Datagram Protocol) – Connectionless.

Computer Networks - Prof Ashok Herur 10

 Transmission Control Protocol (TCP)

• Provides a reliable, end-to-end, byte-stream transmission over an unreliable

.IN
internetwork.

C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 11

 Transmission Control Protocol (TCP)

• Is Connection-oriented

.IN
• Explicit set-up and tear-down of TCP session

C
• Offers Stream-of-bytes service

N
• Receives a stream of bytes from higher layers.

SY
• Passes them as Segments to the lower layers.

U
• Provides Reliable, in-order delivery
• Checksums to detect corrupted data
VT
• Acknowledgments & retransmissions for reliable delivery
• Sequence numbers to detect losses and reorder data

Computer Networks - Prof Ashok Herur 12

TCP header

.IN
Source port Destination port

C
Sequence number
Flags: SYN

N
Acknowledgment
FIN

SY
RST HdrLen 0 Flags Advertised window
PSH
URG

U
Checksum Urgent pointer
ACK VT Options (variable)

Data

Computer Networks - Prof Ashok Herur 13

 TCP Header fields
• Port Number:

.IN
• A port number identifies the endpoint of a connection.
• A pair <IP address, port number> identifies one endpoint of a connection.

C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 14

 TCP Header fields

• Sequence Number:

.IN
• The range of the 32-bit SeqNo is
0 <= SeqNo <= 2 32 -1 ≈ 4.3 Giga

C
• Each sequence number identifies a byte in the byte stream.

N
SY
• Initial Sequence Number (ISN) of a connection is set during connection
establishment.

U
VT

Computer Networks - Prof Ashok Herur 15

TCP Header fields

.IN
C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 16

 TCP Header fields

• Acknowledgement Number:

.IN
• Acknowledgements are piggybacked.
• A segment from A -> B can contain an acknowledgement for a data sent in

C
the B -> A direction.

N
• If a host sends an AckNo in a segment it sets the “ACK flag”

SY
• The AckNo contains the SeqNo of the next byte that a host wants to

U
receive.
• Example: The acknowledgement for a segment with sequence numbers
VT
0-1500 is AckNo=1501.

Computer Networks - Prof Ashok Herur 17

 TCP Header fields

• Header Length:

.IN
• Length of header in terms of 32-bit words (4 bytes)
• The TCP header has a variable length (with minimum 20 bytes)

C
• This 4-bit field can hold numbers from 5 (20 bytes) to 15 (60 bytes)

N
SY
• Flags:
• SYN: Synchronize sequence numbers

U
• Sent in the first packet when initiating a connection
VT
• FIN: Sender is finished with sending
• Used for closing a connection
• Both sides of a connection must send a FIN

Computer Networks - Prof Ashok Herur 18

 TCP Header fields
• Flags: (contd)

.IN
• RST: Causes the receiver to Reset the connection
• Receiver terminates the connection and indicates higher layer

C
application about the reset

N
• ACK: Acknowledgement Number is valid

SY
• URG: Urgent pointer is valid
• If the bit is set, the following bytes, in the range given below,

U
contain an urgent message:VT
SeqNo <= urgent message <= SeqNo+urgent pointer
• PSH: PUSH flag is used to tell the receiver to pass all data (that it currently
has) to the application.
• Normally set by sender when the sender’s buffer is empty.
Computer Networks - Prof Ashok Herur 19
TCP Header fields
• Window Size:

.IN
• Each side advertises the size of its sliding window.
• Window size is the maximum number of bytes that a receiver can accept.

C
• Maximum window size is 216-1= 65535 bytes

N
SY
• TCP Checksum:
• TCP checksum covers over both TCP header and TCP data

U
• Urgent Pointer:
VT
• Only valid if URG flag is set

Computer Networks - Prof Ashok Herur 20

 .IN
C
Establishing a connection (Hand-shaking)

N
SY
U
VT

Computer Networks - Prof Ashok Herur 21

Establishing a TCP Connection: 3-way handshake.
A B
SYN

.IN
CK Each host tells its ISN (Initial
SYN A

C
sequence number) to the other host.

N
ACK

SY
Dat
aD
at

U
a
VT
• Three-way handshake to establish connection
• Host A sends a SYN (open) to the host B
• Host B returns a SYN acknowledgment (SYN ACK)
• Host A sends an ACK to acknowledge the SYN ACK
Computer Networks - Prof Ashok Herur 22
Step 1: A’s Initial SYN Packet

.IN
A’s port B’s port

A’s Initial Sequence Number

C
Flags: SYN
Acknowledgment

N
FIN

SY
RST 20 0 Flags Advertised window
PSH
URG Checksum Urgent pointer

U
ACK
VT Options (variable)

A tells B it wants to open a connection…

Computer Networks - Prof Ashok Herur 23

Step 2: B’s response (SYN-ACK) packet to A

.IN
B’s port A’s port

B’s Initial Sequence Number

C
Flags: SYN

N
A’s ISN plus 1
FIN

SY
RST 20 0 Flags Advertised window
PSH
URG Checksum Urgent pointer

U
ACK VT Options (variable)

B tells A it accepts, and is ready to hear the next byte…

… upon receiving this packet, A can start sending data

Computer Networks - Prof Ashok Herur 24

Step 3: A’s ACK of the SYN-ACK

.IN
A’s port B’s port

Sequence number

C
Flags: SYN

N
B’s ISN plus 1
FIN

SY
RST 20 0 Flags Advertised window
PSH
URG Checksum Urgent pointer

U
ACK VT Options (variable)

A tells B it is now okay to start sending it’s own data (to

A)
… upon receiving this packet, B can start sending
data

Computer Networks - Prof Ashok Herur 25

Tearing down a TCP Connection

A B

.IN
FI
N

C
CK
FIN A

N
At this time, B can still send
data to A

SY
FIN
FIN
ACK

U
VT
• Each end of the data flow must be shut down independently.
• If one end is done, it sends a FIN segment. This means that no
more data will be sent.
Computer Networks - Prof Ashok Herur 26
Error control in TCP

• The Data Link layer implements error control measures like using

.IN
a CRC Checksum and using some ARQ protocols.

C
• If every frame is verified, on EVERY link, at the Data Link layer, can

N
something be wrong, end-to-end, when the message reaches the

SY
destination host?
• While the Link layer checks take care of the errors that may occur

U
on the various links, they cannot take care of the errors that may
VT
occur INSIDE the routers.
• At every router, various fields (TTL, Fragment Offset, etc) will /
may change, leading to a change in the Packet Header checksum.

Computer Networks - Prof Ashok Herur 27

Flow control in TCP

• The Data Link layer implements flow control measures like using

.IN
Sliding window protocols.

C
• What would be the optimum size of a window?

N
• At the Data Link layer (for a link between adjacent nodes), the

SY
window size can be pretty small (1, 2, 4,…) since the Bandwidth-

U
Delay product is low.
VT
• Bandwidth-Delay product broadly indicates how many bits
(segments) could have been transmitted by the time a sender
receives the acknowledgment of a message segment sent earlier.

Computer Networks - Prof Ashok Herur 28

Flow control in TCP

• However, at the Transport Link layer (host to host), the

.IN
Bandwidth-Delay product will be high because of the high delay
required to traverse the entire path to the destination.

C
N
• Therefore, the Window size here (at the Transport Layer) has to

SY
be much larger.
• The same logic applies to the Buffering at the transmitting host.

U
VT
• It has to be much larger than that used at an intermediate
router.

Computer Networks - Prof Ashok Herur 29

Wireless issues in TCP

• Transport layer protocols such as TCP that implement congestion

.IN
control should be independent of the underlying Network and
Data Link layer technologies.

C
N
• But, in practice, there are various issues with wireless links /

SY
networks.
• The main issue is that % packet loss is often used as a measure of

U
congestion. VT
• This % packet loss is normally high on wireless links / networks,
due to the higher rate of transmission error (due to noise).

Computer Networks - Prof Ashok Herur 30

Wireless issues in TCP

• By the time, this % packet loss reaches around 10%, the

.IN
connection is effectively stopped because of the throttling by the
congestion control mechanisms.

C
N
• One solution is to differentiate the packet losses due to

SY
transmission error and due to insufficient bandwidth.
• Things will be more complicated when the path contains a few

U
wired links and a few wireless links.
VT

Computer Networks - Prof Ashok Herur 31

 .IN
C
User Datagram Protocol (UDP)

N
SY
U
VT

Computer Networks - Prof Ashok Herur 32

User Datagram Protocol (UDP)
• The main characteristics of UDP are as follows

.IN
• UDP implements a connectionless transfer.
• It does not require connection establishment prior to data transfer

C
• UDP service is unreliable

N
SY
• UDP does not guarantee the delivery of datagram to the destination
• It does not guarantee the correct order of the datagrams.

U
• UDP computes the checksum for the entire header plus data
VT
• UDP does not buffer the incoming datagrams.
• UDP does not have the feature to segment long messages.

Computer Networks - Prof Ashok Herur 33

User Datagram Protocol (UDP)
• UDP results in speedier communication because it does not spend time

.IN
forming a firm connection with the destination before transferring the data.
• User Datagram Protocol (UDP) is used for:

C
N
• Sending relatively small amounts of data, eliminating concerns regarding

SY
controlling errors or the flow of the packets
• Routing update protocols such as Routing Information Protocol (RIP) and

U
Network control protocols like SNMP. VT
• Multicasting because UDP works well with packet switching.

Computer Networks - Prof Ashok Herur 34

User Datagram Protocol (UDP)
• Real-time applications in which the information needs to be delivered quickly

.IN
and smoothly.
• It is specifically chosen for time-sensitive applications like gaming, playing

C
videos, or Domain Name System (DNS) lookups (convert Domain names to IP

N
addresses).

SY
• For users, it is better to have the overall transmission arrive on time than wait

U
for it to get there in a near-perfect state.
VT
• For this reason, UDP is commonly used in Voice over Internet Protocol
(VoIP) applications as well.

Computer Networks - Prof Ashok Herur 35

UDP header

.IN
0 15, 16
31

C
N
Source Port Number Destination Port Number

SY
U
UDP LengthVT Checksum

Data

Computer Networks - Prof Ashok Herur 36

User Datagram Protocol (UDP)
• Source port number identifies the sender's port, when used, and should be

.IN
assumed to be the port to reply to if needed.

C
• If not used, it should be zero.

N
• If the source host is the server, the port number is likely to be a well-known

SY
port number from 0 to 1023.

U
• If the source host is the client, the port number is likely to be an ephemeral
VT
(temporary, short-lived) port number.
• In networking, you might use well-known ports (0 to 1023), registered ports
(1024 to 49151), and dynamic or ephemeral ports (49152 to 65535), as
defined by the Internet Assigned Numbers Authority (IANA).
Computer Networks - Prof Ashok Herur 37
User Datagram Protocol (UDP)
• Destination port number identifies the receiver's port and is required.

.IN
• UDP Length field specifies the length in bytes of the UDP header and data.

C
• The minimum length is 8 bytes, the length of the header.

N
• The field size sets a theoretical limit of 65,535 bytes (8-byte header +

SY
65,527 bytes of data) for a UDP datagram.
• The Checksum field may be used for error-checking of the header and data (by

U
VT
the Application layer); The field carries all-zeros if unused.

Computer Networks - Prof Ashok Herur 38

 .IN
C
Congestion control

N
SY
U
VT

Computer Networks - Prof Ashok Herur 39

Congestion control
• Congestion control has to ensure that the subnet is able to carry the traffic

.IN
offered by the hosts.
• It is a global issue, involving all hosts, all routers and all links in the

C
network.

N
SY
• When too many packets are present in a small part of the subnet, it leads to
congestion.

U
• Then, the data carrying capacity of the subnet reduces.
VT
• When the number of packets dumped onto the subnet by the hosts is within
the carrying capacity, they are all delivered (except the ones with
transmission errors)
• The number delivered is proportional to the number sent by the hosts
Computer Networks - Prof Ashok Herur 40
Congestion control
• As traffic increases beyond the capacity, the network cannot cope and

.IN
packets are lost.

C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 41

Causes of congestion
• Load greater than the resources in a part of the network

.IN
• More traffic and low bandwidth lines
• Insufficient buffer at the routers

C
N
• Having a very large buffer is counter-productive (why?)

SY
• Slow processors at the routers

U
VT

Computer Networks - Prof Ashok Herur 42

Measures of congestion
• Percentage of packets discarded for lack of buffer space

.IN
• Average queue length at the routers

C
• Average packet delay

N
SY
• Standard deviation of the packet delay

U
• Number of packets that are timed-out and retransmitted
VT
In all the above cases, a rising number indicates growing congestion.

Computer Networks - Prof Ashok Herur 43

Congestion control algorithms
• In spite of the routers adapting to the traffic conditions, and changes in

.IN
topology (adaptive routing), traffic can build up in a part of the network.

C
• The basic principle behind all the Congestion Control algorithms is: Reduce

N
load and / or Increase capacity.

SY
• Often, it is difficult to increase capacity temporarily and at short notice.
• Therefore, most of the algorithms concentrate on reducing (regulating)

U
the load. VT

Computer Networks - Prof Ashok Herur 44

Congestion control algorithms
• Common techniques :

.IN
• Admission Control techniques:
• Leaky bucket algorithm

C
N
• Token bucket algorithm

SY
• Traffic throttling techniques:

U
• Choke packet algorithm VT
• Explicit Congestion Notification

• Load shedding

Computer Networks - Prof Ashok Herur 45

Leaky Bucket algorithm

.IN
C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 46

Leaky Bucket algorithm
• The leaky bucket tries to ensure that the number of packets put out on to the

.IN
network (on an average) is constant, regardless of the burstiness of the input;
It does nothing when input is idle or the rate is below the defined constant

C
rate.

N
• The host injects one packet per clock tick onto the network. This results in a

SY
uniform flow of packets, smoothing out bursts and reducing congestion.

U
• It is implemented as a single-server queue with constant service time.
VT
• If the bucket (buffer) overflows, then packets are discarded.

Computer Networks - Prof Ashok Herur 47

Leaky Bucket algorithm
• When packets are of variable length, it is better to allow a fixed number of

.IN
bytes per tick.

C
• The leaky bucket algorithm imposes the same restriction on all hosts (some

N
sending almost continuously and some sending occasionally).

SY
• In short, it does not allow any burstiness beyond the constant output rate
that has been defined.

U
VT

Computer Networks - Prof Ashok Herur 48

Token Bucket algorithm
• The Token Bucket algorithm allows the output rate to vary, depending on the

.IN
size of the burst.
• In the TB algorithm, the bucket holds tokens.

C
• Each token entitles the host to transmit a fixed number of bytes.

N
• To transmit a packet, the host must capture and destroy the required

SY
number of tokens (depending on the packet size).

U
• Tokens are generated by a clock at the rate of one token every Δt sec.
VT
• Idle hosts can capture and save up tokens (up to the max. size of the bucket)
in order to send larger bursts later.
• When the bucket fills up, new tokens are lost.

Computer Networks - Prof Ashok Herur 49

Token Bucket algorithm

.IN
C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 50

Choke Packet algorithm
• When a router is congested or about to get congested, it starts sending choke

.IN
packets to the sources from where it received packets.

C
• A choke packet is a control packet generated at a congested node.

N
• The source, on receiving the choke packet must reduce its transmission rate

SY
by a certain percentage.

U
• This is a direct way of telling the sources to slow down temporarily.
VT

Computer Networks - Prof Ashok Herur 51

Choke Packet algorithm

.IN
C
N
SY
U
VT

D is the congested router, and A is the source of the congestion

Computer Networks - Prof Ashok Herur 52
Hop-by-hop Choke Packet algorithm
• Over long distances or at high speeds, choke packets are not very effective.

.IN
• A more efficient method is to send to choke packets hop-by-hop.

C
• This requires each hop to reduce its transmission even before the choke

N
packet arrive at the source.

SY
U
VT

Computer Networks - Prof Ashok Herur 53

Hop-by-hop Choke Packet algorithm

.IN
• This technique will spread the
congestion backwards – To the routers

C
along the way.

N
SY
• It is better to have many routers ‘mildly’
congested than having one router

U
VT congested ‘heavily’

Computer Networks - Prof Ashok Herur 54

Explicit Congestion Notification (ECN)
• Instead of generating additional (choke) packets to warn of congestion, a

.IN
router can tag any packet it forwards (by setting a bit in the packet’s header)
to signal that it is experiencing congestion.

C
• When the packet is delivered, the destination notes that there is congestion

N
and informs the sender when it sends a Reply / Acknowledgement packet.

SY
• The sender can then throttle its transmissions, as before.

U
• This method is called Explicit Congestion Notification (ECN).
VT

Computer Networks - Prof Ashok Herur 55

Load shedding
• This method is the last resort, when none of the milder methods, described

.IN
earlier, are able to control congestion.

C
• Load shedding just means that the congested router will throw away packets

N
to reduce the load, when the buffer is filled up / or about to get filled up.

SY
• The key question here is about which packets to discard; The answer depends
on the type of application to which the packets belong.

U
• For a File Transfer, an old packet is worth more than a new one. Dropping packet 6 and
VT
keeping packets 7 to 10 (for example) will only force the receiver to do more work to
buffer the data that it cannot yet use.
• In contrast, in real-time applications, a new packet is worth more than an old one,
because packets become useless if they are delayed and miss the time when they must
be played out.

Computer Networks - Prof Ashok Herur 56

Load shedding – Random Early Detection (R.E.D.)
• Dealing with congestion when it first starts is more effective than dealing

.IN
with it when it gets worse.

C
• Random Early Detection is a type of Load Shedding that discards packets

N
before the buffer space is filled up.

SY
• This idea stems from the fact that most hosts on the Internet do not get
congestion signals in the form of ECN.

U
• The only reliable indication of congestion that they get is Packet loss.
VT
• To determine when to start discarding, routers maintain a running average of
their queue length.

Computer Networks - Prof Ashok Herur 57

Load shedding – Random Early Detection (R.E.D.)
• When the average queue length on some link exceeds a threshold, that

.IN
particular link is said to be congested, and a small fraction of the packets
(coming on that link) are dropped at random.

C
• Picking packets at random makes it more likely that the fastest senders will

N
see more packet drops.

SY
• This is the best option since a router cannot tell which of the sources
(sending packets on that link) is contributing more to the congestion.

U
VT
• The affected sender will notice the loss, and its Transport layer protocol will
slow down.
• The (higher-than-normal) lost packets are thus delivering the same
message as a choke packet, but implicitly.

Computer Networks - Prof Ashok Herur 58

 .IN
C
N
SY
U
VT

Computer Networks - Prof Ashok Herur 59