Abstract

Traffic control cannot be done without having a well-fit model to represent the traffic. Therefore traffic modeling is an essential part of any network control research project. Markov-Modulated Poisson Process (MMPP) shows the great flexibility and analytical tractibility which is needed in traffic control. MMPP model is not only capable of capturing the interframe correlation in the traffic, but also can be easily analysed by using well-known Matrix Geometric techniques. Our research in this project is focused on the study of MMPP for modeling of the traffic in the broadband networks. We first start with the simplest case, a two state MMPP, and study its performance for representing the ATM traffic. Starting with a superposition of voice sources, the performance of various techniques to model the superposed stream by a 2-state MMPP is compared. Then the techniques are generalized for an arbitrary aggregated ATM traffic, characterized only from a sequence of traffic samples. A refined moment-based technique to derive the parameters of a 2-state MMPP model to represent such an arbitrary traffic is proposed. In order to have a more general model applicable to various types of traffic, we propose a special type of multiple-state MMPP, a superposition of N 2-state MMPP minisources. (Abstract shortened by UMI.)