Performance Evaluation of ATM Cell Scheduling Policies Using Stochastic PetriNets

摘要

This paper addresses the modeling and evaluation of cell scheduling policies inasynchronous transfer mode (ATM) switches using stochastic Petri net models. In particular, we compare four cell scheduling policies: a first-in, first out (FIFO) policy with and without non-preemptive priorities, a threshold priority policy, and a modification thereof. The aim of this paper is twofold. First of all, we want to stress the suitability of stochastic Petri nets as a versatile tool for modeling and evaluating ATM-related performance issues. Secondly, we like to point out that under fairly realistic traffic conditions, i.e., a combination of Poisson nonreal-time (data) traffic and Markov modulated Poisson process (MMPP) real-time (video) traffic, our proposed modification of the threshold priority scheduling policy not only reduces the average delay for real-time traffic at the cost of only a slight increase of the average delay of nonreal-time traffic, but also reduces the delay jitter in the real-time traffic.