End-to-End Protocols Overview

End-to-End Protocols

• End-to-end protocols transform host-to-host packet delivery into process-to-process communication.
• Transport protocols provide services like guaranteeing message delivery, preserving sending order, and handling large messages.
• Network limitations impact transport protocols through message dropping, reordering, duplication, and delays.

Chapter Outline

• Simple Demultiplexer (UDP)
• Reliable Byte Stream (TCP)

Chapter Goal

• Understanding demultiplexing services
• Discussing simple byte stream protocols

Common Protocol Properties

• Guarantees message delivery
• Preserves sending order and uniqueness
• Supports arbitrarily large messages
• Enables sender-receiver synchronization.
• Supports multiple application processes on each host, facilitating concurrent communication among many applications.
• Implements flow control (regulating sender data flow to receiver capacity) and congestion control (avoiding network overload).

Network Protocol Limitations

• Message dropping
• Message reordering
• Duplicate message delivery
• Limited message size
• Arbitrary message delay

Challenge for Transport Protocols

• Developing algorithms that translate network limitations into acceptable application level service quality.

Simple Demultiplexer (UDP)

• Extends host-to-host service into a process-to-process communication method.
• Adds demultiplexing layers, allowing multiple applications on a host to share the network, thus promoting concurrent use.
• Provides a connectionless and unreliable service.
• Includes protocols like TFTP, DNS, SNMP, RTP.
• UDP checksum is calculated only if the field is all zeros.

UDP Header Structure

• The header contains source and destination ports, length, and checksum fields.
• Application data follows the header.

Reliable Byte Stream (TCP)

• Provides reliable delivery, unlike UDP's connectionlessness.
• TCP is connection-oriented.
• It offers a byte stream service enabling application processes to treat data as a continuous stream, regardless of underlying packet structure.

Flow Control vs. Congestion Control

• Flow control prevents senders from overwhelming receiver capacity.
• Congestion control prevents overloading the network by regulating data injection.

End-to-End Issues

• TCP runs across the internet, not a single link.
• TCP connections vary in round-trip time (RTT).
• Packets can be reordered.
• Connection mechanisms enable communication partner resource discovery.
• Mechanisms for sender capacity learning are necessary.

TCP Segment

• TCP is byte-oriented.
• Senders write bytes, receivers read bytes from the TCP connection.
• TCP itself transmits data as segments, rather than individual bytes, which is handled by the IP layer.
• Source hosts buffer enough bytes for a packet and send as a transmit segment to destination host.
• Destination hosts empty the packet contents into a buffer, and the receiving process can read data at leisure.
• Segments are units of data exchanged between TCP peers.

TCP Header Format

• TCP header conveys crucial source/destination port details, sequence number, acknowledgment number, window size, and flags (e.g., SYN, FIN, ACK).
• Flexible Options are variable-length.

TCP Header Fields

• SrcPort & DstPort: Identify source and destination ports respectively, crucial for demultiplexing.
• Acknowledgment, SequenceNum, AdvertisedWindow: Key components of the sliding window algorithm.
• Sequence number is assigned to each data byte, facilitating order and acknowledgement of received data.
• Checksum: Data integrity verification mechanism, mirroring UDP's functionality.

TCP Header Flags

• Flags field: Used for communication control between TCP peers.
• SYN, FIN: Flags used for establishing and closing connections respectively.
• ACK: Ensures acknowledgment of received data segments.
• URG: Urgent data transmission indicator.
• PUSH: Pushes data to the receiving application.
• RST: Resets a connection.

TCP Sliding Window

• LastByteAcked, LastByteSent, LastByteWrite, NextByteExpected, and LastByteRcvd.
• Sending side components for ensuring a proper data transmission.
• Receiving side components are also essential for proper data handling and delivery.

TCP Flow Control

• Variables LastByteRcvd, MaxRecvBuffer, AdvertisedWindow, LastByteSent, and effective window parameters control flow, matching receiver capacity via advertised window.
• LastByteWritten, MaxSendBuffer: parameters crucial for efficient data transmission.

Protecting against Wraparound

• SequenceNum uses 32 bits to prevent numbering issues for extensive data transfer windows.
• The AdvertisedWindow field uses 16 bits to specify the receive buffer size.

Keeping the Pipe Full

• The 16-bit field must be large enough for reliable data pipeline operation.
• Receiver-side buffer space and bandwidth delay product interplay for optimizing throughput and addressing performance issues.

Triggering Transmission

• Factors like window size and data availability determine when TCP transmits a segment.
• Application programs dictate data writing input streams.
• TCP decides when sufficient bytes are available to form a transmission.

Nagle's Algorithm

• Optimizes data transmission by waiting for data buffers and acknowledgement packets before transmitting when network conditions are favorable to avoid needless small packets as smaller packets are less efficient to transmit than larger ones.
• Introduces a timer that triggers data transmission if enough bytes are present and no prior data has been sent.

Adaptive Retransmission

• Algorithm dynamically adjusts retransmission timeouts based on measured round-trip times to improve efficiency and responsiveness.
• The exponentially weighted moving average method computes the weighted average by giving more weight to recent observations than older ones.
• Weights help avoid the impact of outliers, improving the robustness of the timeout calculation.

Karn/Partridge Algorithm

• Improves Adaptive Retransmission by not sampling RTT during retransmission, avoiding potential inaccurate RTT estimations, focusing on avoiding inaccuracies when retransmitting packets.
• It avoids sampling RTT during retransmissions to not collect potentially erroneous data.
• It applies a double timeout after each retransmission to improve stability and robustness.
• Improves upon the original algorithm by focusing on eliminating congestion issues and not on getting an RTT directly, which involves avoiding the problems with having multiple retransmissions.

Jacobson/Karels Algorithm

• Addresses the limitations of Karn/Partridge by calculating deviation and adjusting timeouts based on varied RTT samples, enabling more intelligent prediction and less aggressive retransmissions.

TCP, Client/Server

• Highlights the various steps involved in a client-server TCP connection establishment and termination process, providing relevant details via sequence numbers and acknowledgement.

TCP, Connection Termination

• Shows the sequence of steps in TCP connection termination.

This quiz covers end-to-end protocols that facilitate process-to-process communication over host-to-host packet delivery. Topics include UDP demultiplexing and TCP reliable byte stream protocols, alongside their properties like message delivery guarantees and congestion control. Test your understanding of how these transport protocols work to achieve effective data communication.