Polynomial Law: A Better Alternative to Binary Exponential Backoff

In this paper we provide an approximate analysis that synthesizes results on the throughput and the delay of S-Aloha under a variety of backoff laws with an unlimited number of backoff stages. Past results, common to those about IEEE 802.11 protocols, show that, with the exponential backoff, the access-delay distribution is heavy tailed, and that, in order to alleviate the delay, throughput and number of users must be sacrificed. This paper uses an approximated and simple model recently introduced to produce results that are usually derived in a more complex framework. Here the model is extended to general backoff laws, including polynomial laws of type (a + i(nu))(-1). We find that, when considering all performance factors, i.e., throughput, variance, access flexibility, and maximum number of users allowed, the polynomial law is to be preferred, with a degree nu that increases with the number of users that can be accommodated.