Networking Challenges and Reliable Communication

Introduction

• Key challenge in networking: Reliable communication over unreliable channels.
• Channels can lose, corrupt, or reorder messages.
• Aim is to communicate reliably between a sender and a receiver.

Channel Models

• Start with a perfect channel model (no message loss or corruption).
• Gradually introduce realistic assumptions about unreliable channels:
  • Message loss
  • Bit corruption
  • Message reordering

Reliable Data Transfer (RDT) Protocol

• Implemented at the transport layer.
• Unidirectional channel: Sender to receiver focus.
• Bidirectional protocol communication: Sender and receiver sides exchange messages over an unreliable channel.

Considerations

• Channel characteristics affect protocol complexity:
  • Message loss
  • Message corruption
  • Message reordering
• Protocol entities can only know about each other through message exchange.

Interface and API

• Interfaces above and below the transport layer:
  • Sender side: Data is passed from the application layer to the transport layer, becoming a transport layer segment.
  • Receiver side: Transport layer segment is received, data is delivered to the receiving process.

Protocol Specification

• Formal specification needed to avoid misinterpretation.
• Use finite state machines for protocol specification:
  • Notion of state
  • State transitions due to events
  • Actions performed by the system

Example: Light Bulb

• Illustrates state, events, and actions:
  • States: On and off
  • Events: Pressing the on or off switch
  • Actions: Transition between states and light emission

Developing RDT Protocol

• RDT 1.0: Perfect channel, simple finite state machines for sender and receiver.
• RDT 2.0: Deals with bit errors using:
  • Checksums for error detection
  • Acks (Acknowledgements) for correct message receipt
  • Nacks (Negative Acknowledgements) for errors
  • Retransmissions for error recovery

Human Communication Analogy

• Acknowledgements (Yes, I got it)
• Negative acknowledgements (No, resend it)
• Retransmissions (Restate the message)

RDT 2.1 Improvements

• Handles corrupted acks/nacks:
  • Sequence numbers added to packets to detect duplicates.
  • Stop-and-wait protocol with one-bit sequence number.
• Finite state machines expanded: 4 states for different sequence numbers and actions.

RDT 2.1 Receiver

• Waits for specific sequence numbers:
  • Sends acks for correct reception.
  • Sends nacks for corrupted or out-of-sequence packets.

Summary of RDT Protocols

• Use of error detection (checksums), acks/nacks, retransmissions, sequence numbers.
• RDT 2.2: An ack-only version is possible.
• TCP: Relies on acks, sequence numbers, checksums, and retransmissions.

Outstanding Issues

• Addressing packet loss in channels (to be discussed in next section).