Parameterization method for Markov traffic model

Markov modulated models have been frequently used to characterize traffic in communication networks, and such models (MMPP, MMDP, etc) are very suitable for theoretic analysis. However, the parameters of these models are often difficult to obtain. In this paper, we propose and illustrate a new parameterization method directed toward a Markov Modulated Deterministic Process (MMDP) model. Such a model may be used to describe traffic with any distribution. In addition to the theory developed, it is shown by experimental results that our parameterization method works well. We also show that Long Range Dependence (LRD) in video traffic is negligible for many applications. Some experimental results from many traffic traces are presented to show that this kind of simplification is acceptable. We also proposed a new definition of the MMDP model and make a comparison between it and the existing one. It is shown that our definition is more concise and exact. The parameterization method introduced in this paper is based on this new definition.