Karn's Algorithm for Time out Timer

AI Summary

An AI-generated summary of this video lecture.

This lecture segment introduces Karn's Algorithm for TCP timeout timers, focusing on the limitations of basic and Jacobson's algorithms. The instructor explains that both prior methods rely on the actual round-trip time (ARTT) of previous segments to calculate the initial round-trip time (IRTT), using the formula IRTT_{n+1} = α IRTT_n + (1 − α) ARTT_n. A critical problem arises when a segment is retransmitted after its timeout expires: the acknowledgment (ACK) arrives too late to provide a valid ARTT measurement, making standard updates unreliable. The instructor demonstrates this with a numerical example, showing how IRTT values shift from initial estimates (e.g., 10 and 15) to updated figures (12.5 and 25) when invalid data is incorporated. Karn's modification, proposed by Phil Karn and Craig Partridge in 1987, resolves this issue. The core rule is that when a segment is retransmitted due to timeout, the sender must not use the resulting ACK's RTT for estimation; instead, it should double the timeout timer (TOT) to account for potential congestion. This approach prevents skewed IRTT calculations and improves TCP reliability.

Chapters

1. 0:00 – 2:00 00:00-02:00

   The slide titled 'Problems with Basic Algorithm and Jacobson's Algorithm' presents the core issue: both algorithms depend on previous ARTT via IRTT_{n+1} = α IRTT_n + (1 − α) ARTT_n. The instructor uses red pen to underline 'Basic Algorithm' and cross out the α IRTT_n term, emphasizing the dependency problem. A numerical example follows where initial IRTT values of 10 and 15 are crossed out and replaced with 12.5 and 25, illustrating how invalid ACKs after timeout corrupt the estimate. The slide notes that 'Consider ACK of some segment arrives to the sender after its initial time out timer goes off,' highlighting why standard updates fail in retransmission scenarios.

2. 2:00 – 2:57 02:00-02:57

   The lecture transitions to Karn's solution with a slide stating that Karn's algorithm estimates round-trip time for TCP messages, proposed by Phil Karn and Craig Partridge in 1987. Red annotations underline 'Karn' and 'Craig Partridge,' with an arrow pointing to a portrait. The next slide, 'Karn's Modification,' lists the key rule: when a segment is retransmitted (underlined in red), do not use its ACK for RTT estimation. Instead, double the timeout timer (TOT), as emphasized by a red underline and curved stroke beneath 'time out timer (TOT).' Red X marks over 'Basic' and 'Jacobson's' reinforce that this modification supersedes earlier approaches.

The lesson progresses from problem identification to solution. First, it establishes that basic and Jacobson's algorithms fail when ACKs arrive after timeout because they incorporate invalid ARTT data into IRTT calculations. The numerical example concretizes this flaw by showing how estimates drift when corrupted data is used. Karn's modification then provides a clear corrective rule: ignore RTT measurements from retransmitted segments and double the timeout timer instead. This two-part strategy—excluding invalid data and adjusting for congestion—ensures more accurate IRTT estimation. The instructor's red pen annotations (cross-outs, underlines, arrows) visually reinforce these key points, distinguishing the problematic terms in the formula and highlighting Karn's specific modifications. The content is grounded entirely in visible slide text and annotations, with no audio transcript available to confirm additional verbal explanations.