The effect of traffic distribution and transport protocol on WLAN performance

Abstract

Because the statistics of traffic load strongly impact on the queuing and blocking performance of a network, the choice of traffic distribution and transport protocol greatly influences WLAN performance. In this paper we examine the effect of traffic arrival distributions and transport protocols on WLAN performance. In particular, we analyze (by simulation) the effect of four diverse traffic models (Exponential, Pareto, Poisson, and CBR) on the performance of a typical IEEE 802.11 ad hoc network for TCP and UDP. Results obtained show that the network performance for Poisson is almost independent of traffic load for TCP and UDP but not for CBR. However, for both the Pareto and Exponential the network performance is almost independent of load for TCP, but is sensitive to UDP. The network achieves best and worst throughput performance for CBR and Poisson, respectively. Our findings reported in this paper provide some insight into the impact of the choice of traffic models and transport protocols on WLAN performance.