A Simple Markovian Approach to Model Internet Traffic at Edge Routers

In this paper, we propose a simple MMPP (Markov Modulated Poisson Process) traffic model that approximates the LRD (Long Range Dependence) characteristics of traffic traces measured at our institution edge router, at both the flow and packet levels. The MMPP model mimics the real behavior behind the interaction between users, protocols and the network, using the notion of sessions and flows, therefore resulting in a simple and intuitive model. The queueing behavior of the traffic generated by the model is coherent with the one of the measured traces at several different traffic loads. While the model is not intended to offer an explanation of the reasons why Internet traffic is LRD, it does offer a simple and manageable tool for dimensioning and planning networks (link and buffer capacities), since the characteristics of the generated traffic are easily controlled through the model input parameters.