Computer Networks

CSE 301
Course Instructor
DR. HEMA KUMAR YARNAGULA
Assistant Professor
Dept. of Computer Sc and Engg., SRM University - AP
 Unit – IV
INTERNETWORKING AND
TRANSPORT LAYER
Types of Data Deliveries in CN
Types of Data Deliveries in Transport Layer
Role and Position of Transport Layer
Transport Services and Protocols
 Provide logical communication between app processes running on
different hosts

 Transport protocols run in end systems (hosts)

Sender side:
 breaks app messages into segments
 passes segments to network layer

Receiver side:
 reassembles segments into messages
 passes message to application layer
Transport Services and Protocols
 More than one transport protocol available to application
 Internet uses: TCP and UDP, SCTP application
transport
network
data link
physical

application
transport
network
data link
physical

Note:
Network layer: logical communication between hosts
Transport layer: logical communication between processes
IP Addresses versus Port Numbers
Port Numbers
Well Known Port Numbers
Well Known Port Numbers versus Protocols
Internet Transport Layer Protocols
 The transport layer is represented by two protocols: TCP and UDP.

 Reliable, in-order delivery: TCP application

transport
network
 Congestion control data link
physical
network

 Flow control data link

physical
network
data link

 Connection setup physical

network
data link
physicalnetwork
 Unreliable, unordered delivery: UDP data link
physical

 No congestion control network

data link
physical network
application
transport
 No flow control data link
physical
network
data link

 No need for connection setup physical

 Support for broadcast and multicast

UDP: User Datagram Protocol

 UDP stands for User Datagram Protocol.

 UDP is drafted in RFC 768 in August 1980.

 UDP provides a non-sequential transmission of data.

 It is a connectionless transport protocol.

UDP: User Datagram Protocol

 UDP protocol is used in applications where the speed and

size of data transmitted is considered as more important
than the security and reliability.
UDP: User Datagram Protocol
• Provides “best effort” delivery Why still UDP is there?
• No connection establishment
• UDP segments may be: (which can add delay)
• lost
• delivered out of order to app • Simple: no connection state at
sender, receiver
• Connectionless Service: TCP – 20 Bytes, UDP – 8 Bytes
• no handshaking between
UDP sender, receiver • Small segment header
• each UDP segment handled
independently of others • No congestion control: UDP can
blast away as fast as desired
UDP: User Datagram Protocol
UDP: User Datagram Protocol
UDP: User Datagram

 User Datagram is defined as a packet produced by UDP Protocol.

 Each User Datagram has two components:

 A header information
 Data received from the above layer (i.e., Segment)
UDP: User Datagram

 UDP protocol adds the following into the Header:

 Transport level addresses
 Sender Port Number
 Destination Port Number
 Checksum for error control
 Information of length to the data received from the layer above it
UDP: User Datagram Format

Length of UDP segment

(in bytes), including header
UDP: User Datagram Format

Checksum
once again??
UDP: Checksum
Goal: To detect “errors” (e.g., flipped bits) in transmitted Segment.

Sender: Receiver:
• Treat segment contents as • Compute checksum of received
sequence of 16-bit integers segment
• Checksum: addition of segment
contents (1’s complement sum) • Check if computed checksum
equals checksum field value:
• Sender puts checksum value into • NO - error detected
UDP checksum field • YES - no error detected
UDP: Checksum
UDP: Checksum Example
• Three parts of 16 bits each: • The 1’s complement of ‘r’ is:
• 0110011001100110 • 0011010100110101
• 0101010101010101
• 0000111100001111
• At destination, the sum of four
• Adding the three, calling it ‘r’: parts should be:
• 1100101011001010 • 1111111111111111

• Send the four parts: • If the packet is damaged:

• the original three and • 1111101111111111 (zeros!!)
• 1’s complement of ‘r’ to
destination
UDP: Checksum
Why do we provide error checking, again?

 Because, No guarantee that it is provided in all of the links

between source and destination
UDP: Checksum
Applications of UDP
 UDP is considered a suitable protocol for multicasting.

 UDP protocol can be utilized for smaller size data

 UDP is also used for routing updates such as in Routing

Information Protocol (RIP).

 UDP protocol is generally used for real-time applications

Applications of UDP

Some of the networking implementations that use UDP as

its transport layer protocol are:
 BOOTP, DHCP
 DNS (Domain Name Service)
 NNP (Network News Protocol)
 NTP (Network Time Protocol)
 Quote of the day protocol
 TFTP, RTSP, RIP
Advantages of UDP
Disadvantages of UDP
How UDP Works ==>
What if we need
Guaranteed Delivery?
==>
Transmission Control Protocol (TCP): Overview

 TCP is one of the main protocols of the Internet protocol suite.

 TCP is drafted in RFCs: 793, 1122, 1323, 2018, 2581

 It is a connection-oriented protocol for communications

 The Internet Protocol (IP), which establishes the technique for

sending data packets between computers, works with TCP.
TCP: Overview

application application
writes data reads data
socket socket
door door
TCP TCP
send buffer receive buffer
segment
Features of TCP

Point-to-Point (Unicast):
 All communication has one sender, one receiver
Features of TCP
Connection Oriented:
 A secured connection is being established between the sender and
the receiver.

 How: using handshaking (exchange of control msgs)

 It means sender and receiver are connected to each other till the
completion of the data transfer process.

 The order of the data is maintained i.e. order remains same before
and after transmission.
Features of TCP
Full Duplex Connection:
 Bi-directional data flow in same connection

 Data can be transmitted from receiver to the sender or vice - versa

at the same time.

 MSS: maximum segment size

 It increases efficiency of data flow between sender and receiver.

Features of TCP
Flow Control:
 Flow control limits the rate at which a sender transfers data.

 Aim: Sender will not overwhelm receiver

 This is done to ensure reliable delivery.

 The receiver continually hints to the sender on how much data can
be received (using a sliding window)
Features of TCP
Congestion Control:
 TCP takes into account the level of congestion in the network

 Aim: sender will not overwhelm network

 Congestion level is determined by the amount of data sent by a

sender
Features of TCP
Error Control:
 TCP implements an error control mechanism for reliable data
transfer

 Error control is byte-oriented

 Segments are checked for error detection

TCP Segment Structure
 TCP segment contains two things:
 Header Field
 Data Field
TCP: Header Format
TCP: Header Format
 The header of the TCP is of:
 minimum 20 bytes and
 maximum of 60 bytes.

Let us briefly discuss the various fields:

 Source Port: Defines the port address of the source.

 Destination Port: Defines the port address of the destination.

 Sequence Number: It contains the sequence number of a segment or

data bytes in a session.
TCP: Header Format
Acknowledgment Number:
 ACK flags contain next sequence number of the segment of data
 It works as an acknowledgment for the previously received data.

HLEN:
 HLEN stands for Header Length.
 It has 4-bit value.
 Using 4 bits, it specifies the length of the header.
 The minimum value for HLEN is 5 and maximum is 15.

 Reserved: It is reserved for any future use and it is of 6-bits.

TCP: Header Format

Flags:
 The flag is a 1-bit value.
 Flags are of mainly 6 types and are also known as control bits.

1. URG: URG represents an urgent pointer, so, if the URG value is set
to 1 then the data is processed urgently.

2. ACK: ACK denotes acknowledgment, so, if its bit value is set to 0

then the data packet does not contain an acknowledgment.
TCP: Header Format

Flags:
3. PSH: PSH represents PUSH, so, if the PSH bit is set to 1 then the
receiver is requested to push the data without buffering it.

4. RST: RST stands for restart, so if its bit value is set to 1 then the
connection needs to be restarted.
TCP: Header Format

Flags:
5. SYN: SYN bit as discussed earlier is used to establish a connection.

6. FIN: FIN represents finish, so, if its value is set to 1 then the
connection is to be closed.
TCP: Header Format

Window Size:
 The window size is a 16-bit field
 It contains the data size that the receiver can accept.

 It helps in the flow control mechanism.

 The window size field is determined by the receiver only.
TCP: Header Format

Checksum:
 The checksum is also a 16-bit field

 It contains the checksum of the header and the data.

TCP: Header Format

Urgent Pointer:
 If the value of URG flag is set to 1

 Then the urgent Pointer field points to the urgent data byte.
TCP: Header Format

Options and Padding:

 As the name suggests, the option field contains options that are not in the
regular header.

 The options field is described as a 40-byte word.

 Padding is used in the case when the data in the option field is less than 40-
byte
 So, padding (adding extra 0 bits) helps to make the option bit reach the 40-byte
word boundary.
Typical TCP Transaction Client Server

A TCP Transaction consists of 3 Phases:

 Connection Establishment
Connection Establishment
 Handshaking between client and server

 Reliable, In-Order Data Exchange Reliable, In-Order

Data Exchange
 Recover any lost data
 through retransmissions and ACKs
Connection Termination
 Connection Termination
 Closing the connection
time time
TCP Connection Establishment
 TCP sender, receiver establish “connection” before exchanging
data segments

 It initialize TCP variables:

 Seq. Nos, buffers, flow control info (e.g. RcvWindow)

 Client: connection initiator

 Socket clientSocket = new Socket("hostname", port#);

 Server: contacted by client

 Socket connectionSocket = [Link]();
Connection Establishment (contd.)
Three Way Handshake:
 Step 1: client host sends TCP SYN segment to server
 specifies a random initial seq #
 no data is sent.

 Step 2: server host receives SYN, replies with SYNACK segment

 server allocates buffers
 specifies server initial seq. #

 Step 3: client receives SYNACK, replies with ACK segment, which

may contain data
Connection Establishment (contd.)
Three Way Handshake: Host A Host B

Connection request
Host B ACKs and selects
its own initial seq #

Host A ACKs

time time

Three-way handshake
Connection Establishment (contd.)
Three Way Handshake: Host A Host B

Connection request
Seq. #’s:
• byte stream “number” of first Host B ACKs and selects
byte in segment’s data its own initial seq #

ACKs:
• seq # of next byte expected
from other side
Host A ACKs
• cumulative ACK

time time

Three-way handshake
TCP Connection Termination
Closing a connection:
 Client closes socket:
 [Link]();

 Step 1: client end system sends TCP FIN control segment to server

 Step 2: server receives FIN, replies with ACK.

 Server might send some buffered but not sent data before
closing the connection.
 Server then sends FIN and moves to Closing state.
TCP Connection Termination
Closing a connection:
Client Server

close

Data write

timed wait

close

closed
Byte stream to TCP Segment
Emulated Using TCP “Segments”
Host A

Segment sent when:

TCP Data 1. Segment full (Max Segment Size),
2. Not full, but times out, or
3. “Pushed” by application.

TCP Data

Host B
Sequence Numbers in TCP
Host A
ISN (initial sequence number)

Sequence number = 1st byte TCP

TCP Data HDR
ACK sequence number = next expected byte

TCP
TCP Data HDR
Host B
TCP Segment Format
IP Packet:
 No bigger than Maximum Transmission Unit (MTU)
 E.g., up to 1500 bytes on an Ethernet

TCP Packet:
 IP packet with a TCP header and data inside
 TCP header is typically 20 bytes long
IP Data
TCP Data (segment) TCP Hdr IP Hdr

TCP Segment:
 No more than Maximum Segment Size (MSS) bytes
 E.g., up to 1460 consecutive bytes from the stream
TCP Flow Control
Flow Control
Sender won’t overflow receiver’s
buffer by transmitting too much
and too fast

 Speed-matching service: matching the send rate to the receiving

app’s drain rate
TCP Flow Control: How It Works
(Suppose TCP receiver discards out-of-order segments)

 Spare room in buffer

RcvWindow = RcvBuffer - [LastByteRcvd - LastByteRead]
TCP Flow Control: How It Works
 Then Receiver advertises spare room by including value of
RcvWindow in segments

 Sender limits unACKed data to RcvWindow

 This guarantees receive buffer doesn’t overflow

TCP Flow Control:
How It Works
Pipelined Protocols
 Pipelining: sender allows multiple, “in-flight”, yet-to-be-acknowledged
packets
 range of sequence numbers must be increased
 buffering at sender and/or receiver

 Two generic forms of pipelined protocols: Go-Back-N, Selective Repeat

Pipelining -> Increased Utilization
Sender Receiver
first packet bit transmitted, t = 0
last bit transmitted, t = L / R

first packet bit arrives

RTT last packet bit arrives, send ACK
last bit of 2nd packet arrives, send ACK
last bit of 3rd packet arrives, send ACK
ACK arrives, send next
packet, t = RTT + L / R

Increase utilization
by a factor of 3!

3*L/R .024
U = = = 0.0008
sender 30.008
RTT + L / R microsecon

We assume a 1 Gbps link, 15 ms prop. delay, 8000 bit packet:

Pipelined Protocols
Go-back-N: big picture: Selective Repeat: big piccture
• Sender can have up to N • Sender can have up to N unacked
unacked packets in pipeline packets in pipeline
• Rcvr only sends cumulative ACKs • Rcvr acks individual packets
• Doesn’t ack packet if there’s a gap • Sender maintains timer for each
• Sender has timer for oldest unacked packet
unacked packet • When timer expires, retransmit only
• If timer expires, retransmit all unack packet
unacked packets
Congestion Control

Congestion:
 informally: “too many sources sending too much data too fast for
network to handle”.

 It is different from flow control!!

 Its Manifestations:
 lost packets (buffer overflow at routers)
 long delays (queueing in router buffers)
Causes/costs of congestion
 Every router has finite buffers
 The sender retransmission of lost packet
Host A lin : original data lout

l'in : original data, plus

retransmitted data
Host B finite shared output
link buffers
Causes and costs of congestion
Other factors could be:
 Excessive consumption of bandwidth
 Improper subnet management
 Multicasting
 Outdated hardware
 Aggressive routing protocol
Approaches towards Congestion Control

End-end congestion control Network-assisted congestion control

Open Loop Congestion Control

 Open loop congestion control policies are applied to prevent

congestion before it happens.

 Also called as Congestion-Avoidance

 The congestion control is handled either by the source or the

destination.

 This is implemented in the TCP Layer

Closed Loop Congestion Control

 Closed loop congestion control techniques are used to treat

or alleviate congestion after it happens.

 This is implemented in the Network Layer

Congestion Control
End-end vs Network-assisted

End-end congestion control: Network-assisted congestion

control:
• no explicit feedback from
network • routers provide feedback to end
systems (i.e., to senders)
• congestion inferred from end-
• single bit indicating congestion
system observed packet loss,
(ECN)
packet delay
• explicit rate sender should send
• approach taken by TCP at
TCP Congestion Control

 TCP congestion control is a method used by the TCP protocol

 Why: To manage data flow over a network and prevent/avoid

congestion.

 How: TCP uses a congestion window and congestion policy that

avoids congestion.
TCP Congestion Control

 Previously, we assumed that only the receiver could dictate the

sender’s window size.

 We ignored another entity here, the network.

 What if the network cannot deliver the data as fast as it is created

by the sender?

 In other words, the network is a second entity that also

determines the size of the sender’s window.
Congestion Policy in TCP

 Congestion in TCP is handled by using these three phases:

 Slow Start Phase:
 Congestion Avoidance Phase:
 Congestion Detection Phase:
TCP Congestion Control
Different Variables used in TCP Congestion Control
Congestion Window (Cwnd)
 It limits the amount of data to be sent by the sender into the
network even before receiving the acknowledgment.

Slow start Threshold (ssThresh)

 It is a value, which is used in the slow start phase

Maximum Receiver Capacity / Receiver Window Size

 The sender should not send data greater than that of the size of
receiver window.
Slow Start Phase

 Initially, sender sets congestion window size = maximum segment

size (1 MSS)
CWnd = 1 MSS

 The sender increases the size of the congestion window (Cwnd)

after receiving the ACK (acknowledgment).

 The size of the congestion window increases exponentially in

this phase.
Slow Start Phase
Slow Start Phase
 This phase continues until Cwnd reaches its slow start threshold.
Slow Start Phase
 How this slow start threshold value is computed:
Congestion Avoidance Phase

After the SS_threshold is reached, it enters Avoidance phase:

 The size of the congestion window is increased by the sender
linearly

 Why: in order to avoid congestion.

 Each time an acknowledgment is received, the sender increments

the size of the congestion window by 1.
Congestion Avoidance Phase
Congestion Avoidance Phase

 This phase continues until the size of the CWnd window

becomes equal to that of the receiver window size.

 Then, it is set to constant to avoid overwhelming receiver.

Congestion Detection Phase

 However, during this the sender can identify the segment loss

 The sender can act depending on the type of loss detected.

 Case-01: Detection On Time Out
 Case-02: Detection Of Receiving 3 Duplicate Acknowledgements
Congestion Detection Phase
Case-01: Detection On Time Out
 In this, the time-out timer expires even before receiving
acknowledgment for a segment.

 It suggests a stronger possibility of congestion in a network

 there are chances that a segment has been dropped in the network
Congestion Detection Phase
Reaction in response to Detection on time out:
 Setting the threshold to start at half of the current size of the
window
i.e., SS_threshold = CWnd/2

 Decreasing the size of the congestion window to 1 MSS

 Slow start phase is resumed

Congestion Policy in TCP
Congestion Detection Phase
Case-02: Detection Of Receiving 3 Duplicate Acknowledgements

 This case suggests the weaker possibility of congestion in the

network.

 The chances are that fewer segments have dropped while the one
sent later might have reached.
Congestion Detection Phase
Reaction on receiving 3 duplicate acknowledgments:

 Setting the threshold to start at half of the current size of the

window
i.e., SS_threshold = CWnd/2

 Decreasing the size of the congestion window to that of the slow

start threshold

 The congestion avoidance phase is resumed

Congestion Policy in TCP
Congestion Policy in TCP
Congestion Policy in TCP
 Incrementing the Cwnd by 1 MSS (for each ACK received) and
reducing SS_Threshold to half of the current size of the window (on
segment loss) is called as:

TCP’s Additive Increase Multiplicative Decrease (AIMD)