TCP Reliable Transmission Mechanism Explained

TCP reliable transmission is a core feature of the TCP protocol, ensuring data can be transmitted correctly, orderly, and without loss in unreliable network environments.

Mechanisms Ensuring Reliable Transmission

1. Sequence Number and Acknowledgment

• Sequence Number: Each byte has a unique sequence number, identifying its position in the stream
• Acknowledgment (ACK): Receiver sends ACK after receiving data, confirming received data
• Cumulative Acknowledgment: ACK number indicates the next byte sequence number expected to receive
• Purpose: Ensure data arrives in order, detect lost packets

2. Retransmission Mechanism

Timeout Retransmission (RTO)

• RTO Calculation: Dynamically calculated based on RTT (Round Trip Time), typically RTO = RTT + 4 × RTT variance
• Trigger Condition: No ACK received within RTO time after sending data
• Exponential Backoff: After each retransmission, RTO doubles, avoiding frequent retransmissions during network congestion

Fast Retransmit

• Trigger Condition: Receive 3 duplicate ACKs
• Advantage: Retransmit lost segment immediately without waiting for RTO to expire
• Efficiency: Recover data transmission faster than timeout retransmission

3. Checksum

• Calculation Range: TCP header, data, and pseudo-header (including IP addresses, etc.)
• Purpose: Detect errors during data transmission
• Handling: Discard segment directly on checksum error, do not send ACK

4. Data Ordering

• Problem: Packets may arrive out of order in the network
• Solution: Receiver reorders based on sequence numbers
• Caching: Out-of-order data is cached first, waiting for missing data to arrive before delivering to application layer together

5. Flow Control

• Sliding Window: Receiver advertises available buffer size
• Prevent Overflow: Avoid sender sending too fast causing receiver buffer overflow

6. Congestion Control

• Slow Start: Gradually increase sending rate
• Congestion Avoidance: Linear growth, avoid network congestion
• Fast Retransmit and Fast Recovery: Quickly respond to packet loss, recover transmission

Trade-offs of Reliable Transmission

• Latency vs Reliability: Retransmission mechanism increases latency but improves reliability
• Efficiency vs Safety: Checksum adds overhead but ensures data integrity
• Throughput vs Stability: Congestion control reduces throughput but maintains network stability

Related Questions

• How does TCP handle out-of-order arriving packets?
• How is RTO dynamically calculated and adjusted?
• Why do we need fast retransmit mechanism?