Reliable Data Transfer RDT2.1 Overview

Reliable Data Transfer - RDT2.1

• Problem: RDT1.0 assumes a perfect channel. RDT2.0 introduces a channel with bit errors.
• Solution: Checksums detect errors, but a mechanism is needed for the receiver to inform the sender about problems.
• Methods:
  • ACK (Acknowledgement): The receiver confirms successful packet reception.
  • NAK (Negative Acknowledgement): The receiver indicates a packet error.
• Sender: The sender retransmits a packet if it receives a NAK.
• Receiver: The receiver sends an ACK if the packet is received correctly and a NAK if there’s an error.

RDT2.1 - Sender FSM

• Initial State: The sender starts with a sequence number of 0.
• Event 1: When rdt_send is called, the sender creates a packet with data and a checksum.
• Action 1: The packet is sent using udt_send.
• Event 2: When an ACK or NAK arrives:
  • If corrupt, the sender retransmits.
  • If it's an ACK for sequence number 0, the sender moves to state 1.
• Event 3: When rdt_send is called again, the sender creates a packet with data and a checksum.
• Action 3: The packet is sent using udt_send and the sender moves to state 1.

RDT2.1 - Receiver FSM

• Initial State: The receiver is expecting a packet with sequence number 0.
• Event 1: When a packet arrives:
  • If corrupted or doesn’t have the correct sequence number, send a NAK.
  • If the packet is valid and has sequence number 0, extract the data and deliver it. Send an ACK for sequence number 0.
• Event 2: When the receiver receives an ACK for sequence number 0, it moves to the state where it is expecting sequence number 1.

Handling Duplicate Packets

• Solution: The sender adds a sequence number to each packet.
• Receiver: The receiver discards duplicate packets.

RDT2.1 - Discussion

• Sender: Adds a sequence number to packets.
• Receiver: Checks the sequence number to identify duplicate packets.

Pipelined Protocols

• Go-Back-N: Allows multiple unacknowledged packets to be in transit. The receiver only sends cumulative ACKs (acknowledging all packets up to a certain sequence number).
• Selective Repeat: Allows multiple unacknowledged packets to be in transit. The receiver sends an individual ACK for each packet.

Go-Back-N

• Sender:
  • Uses a k-bit sequence number in the packet header.
  • Maintains a "window" of up to N unacknowledged packets.
  • Sends packets in sequence.
  • If the timer for the oldest unacknowledged packet expires, retransmit all unacknowledged packets.
• Receiver:
  • Only accepts packets in sequence.
  • Sends cumulative ACKs.
  • Discards out-of-order packets.

Selective Repeat

• Sender:
  • Maintains a timer for each unacknowledged packet.
  • If the timer expires, retransmit only the corresponding packet.
• Receiver:
  • Can buffer out-of-order packets.
  • Sends individual ACKs for each packet.

TCP Congestion Control

• Problem: Sending too many packets can overwhelm the network and cause congestion.
• Goal: Optimize the sending rate to avoid congestion.
• Mechanism:
  • Congestion Window (cwnd): Limits the amount of data the sender can send at a time.
  • Slow Start: Initially starts with a small cwnd and increases it exponentially until a loss event.
  • Congestion Avoidance (CA): Increases the cwnd linearly after a loss event.
• Loss Detection:
  • Timeout: A packet hasn't been acknowledged within a certain timeframe.
  • 3 Duplicate ACKs: Indicates packet loss.
• Reaction to Loss:
  • Timeout: cwnd is set to 1.
  • 3 Duplicate ACKs: cwnd is cut in half.
• AIMD (Additive Increase Multiplicative Decrease): Progressively probes for available bandwidth and adjusts the cwnd accordingly.
• TCP Reno: A specific implementation of congestion control that uses 3 duplicate ACKs to detect loss.

TCP - Switching from Slow Start to CA

• Condition: When cwnd reaches half its value before the timeout occurred.
• Reason: Indicates that the network can handle a larger amount of data.