Computing TCP's Retransmission Timer
By Vern Paxson, Mark Allman. Request for Comments, 2000
1,087 citations recorded by the source corpus.
This document defines the standard algorithm that Transmission Control Protocol (TCP) senders are required to use to compute and manage their retransmission timer. It expands on the discussion in section 4.2.3.1 of RFC 1122 and upgrades the requirement of supporting the algorithm from a SHOULD to a MUST.
Subfields: Computer Networks and Communications, Hardware and Architecture
View Paper: https://www.semanticscholar.org/paper/e25656a27403628fc311a47cc9d299e7a78fb549
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Networks and Communications
- Jean-Pierre Hubaux
- Carlo Caini
- Rosario Firrincieli
- Subhabrata Sen
- Imad Aad
- Oliver Spatscheck
- Edward W. Knightly
- Yunhong Gu
In Hardware and Architecture
- Dan Wing
- Philip Matthews
- Oliver Spatscheck
- G. R. Ganger
- Aleksandar Kuzmanovic
- Robert L. Grossman
- Edward W. Knightly
- Jean-Pierre Hubaux
Citations in the corpus
- An in-depth study of LTE: effect of network protocol and application behavior on performance (2013, methodology): “Although it is recommended that the RTO should be at least 1 second [20], depending on the operating systems, different minimum values are used, e.g., Linux’s minimum RTO is 200ms [28].”
- Session Traversal Utilities for NAT (STUN) (2020, methodology): “First, the initial value for RTO SHOULD be configurable (rather than the 3 s recommended in RFC 2988 ) and SHOULD be greater than 500 ms. The exception cases for this "SHOULD" are when other mechanisms are used to derive congestion thresholds (such as the ones defined in ICE for fixed rate streams)”
- TCP Hybla: a TCP enhancement for heterogeneous networks (2004, background): “To this end it is necessary either to wait the ACK of the first non-retransmitted packet or to make use of timestamps [24].”
- SOAR: Simple Opportunistic Adaptive Routing Protocol for Wireless Mesh Networks (2009, methodology): “Then the node updates its RT O as shown in Figure 5. RT O is initialized to 30 ms. Our evaluation uses K = 4, ® = 1/8, and ¯ = 1/4 as in TCP [ 24 ].”
- UDT: UDP-based data transfer for high-speed wide area networks (2007, methodology): “The timeout value can be assigned by users, otherwise it uses the default value according to the TCP RTO calculation described in RFC 2988 [51].”
- The NewReno Modification to TCP's Fast Recovery Algorithm (2004, background): “We note that the Impatient variant in Section 3 doesn’t follow the recommended algorithm in RFC 2988 of restarting the retransmit timer after every packet transmission or retransmission [ RFC2988 , Step 5.1].”
- Denial of service resilience in ad hoc networks (2004, background): “The study showed that the impact of the attack can be quite severe whether minimum RTO values are all set to 1 second as recommended in [31], or are randomized over a wide range.”
- Low-rate TCP-targeted denial of service attacks and counter strategies (2003, methodology): “Moreover, if all flows have an identical minRTO parameter as r ecommended in RFC 2988 [28], the TCP flows can be forced into continual timeouts if an attacker creates periodic outages .”
- Measurement and Analysis of TCP Throughput Collapse in Cluster-based Storage Systems (2008, methodology): “Implementation Problems: Reducing RTOmin to 200μs requires a TCP clock granularity of 100μs, according the standard RTO estimation algorithm [29, 19].”
- this paper (2009, background): “Due to the highly skewed distribution of decoding delay, existing outlier removal mechanisms based on standard deviation (e.g., the three-σ rule [65, 68]) suffer from differentiating stationary state transitions from outliers.”