MMPP modeling of aggregated ATM traffic

Abstract

The paper presents a study on the use of Markov-modulated Poisson processes (MMPPs) in modeling aggregate ATM traffic in order to evaluate the queueing performance. Performance of various techniques for matching a 2-state MMPP to a superposition of on-off sources is compared. The applicability of the 2-state MMPP and its corresponding parameter matching techniques to model an arbitrary aggregate ATM traffic in evaluating its queueing performance are examined by simulation. A refined matching technique for a 2-state MMPP model is proposed. A procedure to estimate the four parameters of a 2-state MMPP from the measured arrivals suitable for the moment-based matching technique is presented. Test cases are given to illustrate the accuracy of the proposed estimation procedure. An approximation for the survivor function is derived.