Computing TCP's Retransmission Timer

Computing TCP's Retransmission Timer

June 2011 | V. Paxson, M. Allman, J. Chu, M. Sargent
This document defines the standard algorithm for computing and managing TCP retransmission timers. It updates RFC 1122 by changing the requirement from a SHOULD to a MUST and obsoletes RFC 2988. The algorithm calculates the retransmission timeout (RTO) using smoothed round-trip time (SRTT) and round-trip time variation (RTTVAR). The initial RTO is set to 1 second, unless the SYN or ACK of the SYN is lost, in which case it reverts to 3 seconds. The RTO is computed as SRTT plus the maximum of G (clock granularity) and K (4) times RTTVAR. The algorithm also includes rules for handling RTT samples, clock granularity, and managing the retransmission timer. The RTO is adjusted based on the received ACKs and may be doubled upon retransmission. The document also discusses security considerations and changes from RFC 2988, including the rationale for lowering the initial RTO. The analysis of various datasets supports the use of a 1-second initial RTO, showing performance benefits for many connections while minimizing negative impacts. The authors are Vern Paxson, Mark Allman, H.K. Jerry Chu, and Matt Sargent.This document defines the standard algorithm for computing and managing TCP retransmission timers. It updates RFC 1122 by changing the requirement from a SHOULD to a MUST and obsoletes RFC 2988. The algorithm calculates the retransmission timeout (RTO) using smoothed round-trip time (SRTT) and round-trip time variation (RTTVAR). The initial RTO is set to 1 second, unless the SYN or ACK of the SYN is lost, in which case it reverts to 3 seconds. The RTO is computed as SRTT plus the maximum of G (clock granularity) and K (4) times RTTVAR. The algorithm also includes rules for handling RTT samples, clock granularity, and managing the retransmission timer. The RTO is adjusted based on the received ACKs and may be doubled upon retransmission. The document also discusses security considerations and changes from RFC 2988, including the rationale for lowering the initial RTO. The analysis of various datasets supports the use of a 1-second initial RTO, showing performance benefits for many connections while minimizing negative impacts. The authors are Vern Paxson, Mark Allman, H.K. Jerry Chu, and Matt Sargent.
Reach us at info@futurestudyspace.com
Understanding Computing TCP's Retransmission Timer