Effectiveness and issues of congestion control in 802.11g wireless LANs

Abstract

The current paper presents a collection of experimental data portraying the performance achieved in the wireless setting by several TCP-friendly congestion controls recently proposed in literature. This work is partly motivated by the consideration that the majority of the analytical results in this area are validated by simulation, rather than by field tests. Examining these algorithms in real environments can help verify their actual effectiveness over the wireless Internet. To reach such goal, two representative controls among the so-called window-based TCP-friendly schemes have been implemented, namely, the General Additive Increase Multiplicative Decrease (GAIMD) strategy, and the SQuare RooT (SQRT) binomial control; the most representative algorithm among rate-based controls, the TCP Friendly Rate Control (TFRC), has also been considered. Their TCP-fairness and smoothness have been comprehensively evaluated in an IEEE 802.11g Wireless Local Area Network (WLAN). The obtained results show that the GAIMD and SQRT strategies reveal non-negligible scalability and smoothness problems, that markedly limit their performance. It is empirically demonstrated that their "optimal" increase/decrease rules, based on TCP-Reno analytical model, cannot guarantee an adequate performance when GAIMD and SQRT compete with TCP-Sack, a de facto standard for current TCP implementations. TFRC is demonstrated to occasionally behave bewildering: properly tuning one of its congestion control parameters and enhancing the algorithm with a flow-control mechanism result in a definitely fairer share of bandwidth with concurrent TCP flows.