CS2005 Networks and Operating Systems - Transport Layer II

Page 1: Introduction

• Title: CS2005 Networks and Operating Systems Transport Layer II

• Authors: J.F Kurose and K.W. Ross

• Copyright: ©1996-2016

Page 2: Session Objectives

• Understand Key Concepts:

  • Connection-oriented transport

  • Transmission Control Protocol (TCP)

  • TCP Segment Structure

  • Reliable data transfer

  • Flow control

  • Connection management

  • Congestion control

Page 3: Overview of Transport Layer Protocols

• Transport layer provides logical communication between application processes on different hosts.

• Implemented in end systems, not in network routers.

• Two key transport protocols in the Internet:

  • TCP (Transmission Control Protocol)

  • UDP (User Datagram Protocol)

• Services provided by transport layer to applications:

  • Process-Level Addressing

  • Multiplexing and Demultiplexing

  • Integrity Checking and Error Detection

  • Connection Management (TCP only)

  • Acknowledgments and Retransmissions

  • Flow Control

• Socket Identification:

  • UDP socket: identified by a 2-tuple (destination IP address + destination port number).

  • TCP socket: identified by a 4-tuple (source IP address + source port number + destination IP address + destination port number).

Page 4: Connection-Oriented Transport (TCP)

• TCP requires a handshake between two processes before data transfer begins.

• TCP connection is logical; state resides only in the end systems and not in intermediate routers.

• TCP provides:

  • Full-duplex service: simultaneous data flow in both directions.

  • Point-to-point connection: between a single sender and a single receiver.

• The connection establishment follows a three-way handshake process:

  1. Client sends a TCP segment to establish a connection.

  2. Server responds with a connection acknowledgment segment.

  3. Client acknowledges the server’s connection acknowledgment.

Page 5: TCP Connection Details

• A TCP connection has buffers, variables, and socket connections on both ends without allocation in network elements.

• Data stream is passed through TCP’s send buffer, and segments sent to the network layer.

• Maximum Segment Size (MSS) is determined by link layer frame size to ensure TCP segments fit into a single frame.

• Each data chunk is paired with a TCP header to form TCP segments, which are encapsulated in IP datagrams.

• Both sender and receiver maintain their own send and receive buffers.

Page 6: TCP Segment Structure

• TCP segment fields: source and destination ports, checksum, sequence number, acknowledgment number, receive window, header length.

• The sequence number identifies the byte-stream number of the segment's first byte.

• Acknowledgment number represents the next byte the receiver expects.

• The TCP header is typically 20 bytes long, allowing for an optional field for maximum segment size and other parameters.

• Header flags:

  • ACK: acknowledgment segment.

  • SYN: connection initiation.

  • FIN: connection termination.

  • CWR, ECE: congestion notification flags.

  • PSH: push data to the upper layer immediately.

  • URG: urgent data indication.

Page 7: Sequence and Acknowledgment Numbers

• Sequence Number Field

  • First byte's number in the segment.

• Acknowledgment Number Field

  • Indicates the next byte expected from sender.

• Example of sequence/acknowledgment with data segments and their corresponding numbers.

Page 8: Sequence and Acknowledgment Example

• Client sends 'C' with seq 42 to server and receives 'C' back.

  • Client's segments:

    • Seq 42, ACK 79, payload 'C'.

    • Seq 43, ACK 80.

  • Server's segment:

    • Seq 79, ACK 43, payload 'C'.

Page 9: TCP Reliable Data Transfer

• TCP ensures data stream integrity; data is uncorrupted, without gaps, duplication, or disorder.

• Utilizes timeouts to manage lost segments and utilizes cumulative acknowledgments.

• Potential issues with timeout settings affecting retransmissions.

Page 10: Fast Retransmit

• Duplicate ACKs used to identify lost segments before timeouts occur.

• Upon receiving three duplicate ACKs, fast retransmit is triggered for the presumed lost packet.

Page 11: TCP Flow Control

• TCP uses flow control to match sending rates with receiving application rates.

• Prevents overflow in the receiving buffer by utilizing a variable called the receive window (rwnd).

• The receive window's size must always be managed within the limits of the receive buffer size.

Page 12: TCP Connection Management: Establishment

• Connection establishment involves a three-step handshake:

  1. Client sends SYN segment with chosen initial sequence number.

  2. Server acknowledges with SYN-ACK, allocates buffers, and sets initial sequence.

  3. Client acknowledges server's SYN-ACK.

Page 13: TCP Connection Management: Termination

• Either process can terminate a connection. Resources are deallocated upon closure.

• Closure process entails following a four-segment exchange using FIN and ACK flags.

Page 14: TCP Congestion Control

• Focuses on managing withdrawal from network congestion using a congestion window (cwnd).

• Traffic sending rate is adjusted based on perceived congestion.

• TCP uses an additive increase/multiplicative decrease method for managing sending rates.

Page 15: Why Use UDP?

• UDP allows application-level control, quick data transmission without connection delays, and lower overhead.

• Suitable for real-time applications that can tolerate some data loss.

• Comparatively, TCP introduces delays due to connection establishment and tracks connection state, affecting performance with many clients.

Page 16: Session Objectives (Repeat)

• Understand Key Concepts from the session about TCP

Page 17: Reading Topics

• Topics covering TCP include:

  • TCP Connection

  • TCP Segment Structure

  • Reliable Data Transfer

  • Flow Control

  • Connection Management

  • Congestion Control