Performance Optimization of CSMA Networks With a Finite Retry Limit

Abstract

A retry limit is usually adopted in practical carrier sense multiple access (CSMA) networks, where a packet is discarded if the maximum number of retransmission attempts is reached. Despite extensive studies, the effect of retry limit on the performance optimization of CSMA networks has remained largely unknown. This paper focuses on a CSMA network with a finite retry limit $M$ , and aims to address the following open issues. First, for a given retry limit $M$ , how should the backoff parameters be adaptively tuned to achieve the optimal network performance? Second, how does the optimal network performance vary with $M$ ? Specifically, in this paper, the explicit expressions of the network steady-state points, the network throughput, and moments of access delay of successfully transmitted packets are all obtained as the functions of the retry limit $M$ , based on which the optimal network performance is further characterized. It is revealed that a CSMA network with a finite retry limit $M$ may have three steady-state points, and the retry limit $M$ has distinct effects on the throughput and delay performance at these steady-state points. The maximum network throughput is found to be independent of $M$ . Yet to achieve the maximum network throughput, the aggregate input rate and the initial transmission probability of each node should be set according to $M$ in unsaturated and saturated conditions, respectively. To optimize the mean access delay, on the other hand, the initial transmission probability should be carefully selected in saturated conditions. The minimum mean access delay can be greatly reduced by choosing a smaller retry limit $M$ , which, nevertheless, leads to a significant throughput loss. The analysis sheds important light on performance optimization of practical CSMA-based networks such as IEEE 802.11 networks.