Statistical multiplexing of aggregate data traffic in ATM networks

Abstract

Aggregate data traffic generated by LAN interconnection over ATM service will certainly be a major component of ATM network traffic. Therefore, it is important to correctly dimension ATM networks in order to support this kind of traffic. In order to achieve this goal, we need a precise traffic model. Bellcore researchers, based on rigorous statistical analyses of a great amount of aggregate Ethernet traffic, have proposed the self-similar or fractal model for this kind of traffic. This model has very distinct statistical properties when compared to traditional aggregate data traffic models such as the Poisson, and the exponential ON-OFF model. Additionally the pseudo-self-similar model, a Markovian model that behaves close to self-similar traffic, has been proposed. This model aims at easing the performance evaluation of ATM networks with self-similar traffic by re-using the analytic methods developed for Markovian models. We study the statistical multiplexing of aggregate data traffic in ATM networks. We aim at determining the impact of the self-similar model in statistical multiplexing (with respect to the Poisson and exponential ON-OFF models) and the feasibility of using the pseudo-self-similar model in this kind of study.