Karn's Algorithm For Optimizing TCP

Karn’s approach addresses the challenge of obtaining accurate round-trip time (RTT) estimates in Transmission Control Protocol (TCP) communications. Proposed by Phil Karn and Craig Partridge in 1987, the method is widely known as the Karn-Partridge Algorithm.

Karn's Algorithm in TCP

In TCP, accurately estimating the Round-Trip Time (RTT) is essential for determining the appropriate Retransmission Timeout (RTO). RTT is generally measured as the time elapsed between sending a segment and receiving its acknowledgment (ACK). However, when a segment is retransmitted, ambiguity arises the sender cannot tell whether the ACK is for the original transmission or the retransmission. This uncertainty leads to unreliable RTT measurements.

Karn’s Algorithm addresses this issue by excluding retransmitted segments from RTT calculations. It updates the RTT estimate only when acknowledgments are clearly for segments that were sent exactly once, eliminating ambiguity and improving accuracy in timing estimations.

Despite its effectiveness, the algorithm has limitations. If there’s a sudden increase in network delay, and TCP relies on outdated RTT estimates, the RTO may become too short. This can cause TCP to retransmit segments unnecessarily, never obtaining a valid RTT sample, and resulting in a cycle of repeated retransmissions and reduced performance.

Working of Karn's Algorithm

• Handling Retransmission Ambiguity: When a packet is retransmitted and an acknowledgment (ACK) is received, TCP cannot determine if the ACK is for the original or retransmitted packet. This ambiguity makes the RTT measurement unreliable.

• Ignoring RTT for Retransmitted Segments: To prevent incorrect RTT estimates, Karn’s Algorithm ignores RTT samples from retransmitted packets. Only ACKs for packets sent once are used to update RTT.

• Exponential Backoff (RTO Adjustment): TCP uses an exponential backoff strategy, doubling the Retransmission Timeout (RTO) after each failed transmission. This helps avoid flooding the network during congestion.

• Updating RTT After Successful Transmission: When a packet is successfully acknowledged without retransmission, its RTT is considered valid and used to update the Smoothed RTT (SRTT) and RTO values.

• Preventing Network Overload: By combining RTT filtering and RTO backoff, Karn’s Algorithm maintains accurate timing and prevents the network from being overwhelmed by repeated retransmissions.

Steps Involved :

1. Send a segment and start a timer.
2. If ACK is received without retransmission, use the time taken to update RTT and RTO.
3. If retransmission occurs, do not update RTT from that ACK.
4. Double the timeout value (RTO) after each retransmission to reduce further congestion.

Need of Karn's Algorithm

• Improves TCP performance in unreliable or varying network conditions.
• Monitors RTT and RTO to help detect and diagnose network problems early.
• Addresses retransmission ambiguity when an ACK is received, it’s unclear if it's for the original or retransmitted packet.
• Inaccurate RTT estimation can lead to incorrect RTO values and inefficient retransmissions.
• Ambiguous ACKs are ignored, ensuring only reliable data is used for RTT updates.
• Prevents repeated unnecessary retransmissions, improving network efficiency and application responsiveness.
• Enables better network tuning and stability, especially in lossy or congested environments.