Performance evaluation of TCP congestion control algorithms using a network simulator

Abstract

The algorithms used in Transmission Control Protocol (TCP) congestion control are the main reason for the performance of a given TCP variant. There are different implementations among TCP: Tahoe, Reno, NewReno, SACK and Vegas. Simulation was used to evaluate these TCP congestion control algorithms from many aspects, such as effective resource utilization (throughput utilization and packet-dropping probability). A network simulator was used with different scenarios and factors to influence the performance of the TCP variants. Under congested network conditions, the more conservative algorithm of TCP Vegas achieves higher throughput and a lower dropped-packet rate than the more aggressive algorithms of the other TCP variants tested.