Analytical Model of the TCP Reno Congestion Control Procedure through a Discrete-Time Markov Chain

Abstract

An analytical model of the Reno congestion control procedure for Transmission Control Protocol is presented, and its theoretical predictions are compared with real TCP traces and some well-known results [1]. The model is based on Discrete-Time Markov Chain, and it covers slow start, congestion avoidance, fast recovery, fast retransmit, cumulative and selective acknowledgments, timeouts with exponential back-off and appropriate byte counting features of TCP. The model provides a way to estimate Reno TCP performance as a function of round trip time and loss rate for bulk transfer TCP flow. The model allows to compare SACK performance with traditional cumulative acknowledgments. The model can be used to find conditions beneficial for advertised window adjustment to improve performance. © Springer International Publishing Switzerland 2014.