A Markovian approach with batch processes for multifractal network traffic

For many real network traffic scenarios, multifractal behavior has been seen to appear within short time scales. It is indicated by fluctuation of the Hurst parameter across small time scales, which directly implies fluctuations in the degree of traffic burstiness in that time scales. In this paper, we present a model for network traffic based on the Markovian-Modulated Batch Poisson Process (MMBPP). The batch length is distributed according to the discrete Pareto distribution in order to reflect high variability (burstinesss) of the traffic across short time scales. The preference for using the MMBPP model is motivated by the fact that it can capture both the variability and correlation of traffic, while at the same time it is expected to be analytically tractable. Numerical results show that the proposed model is able to provide a good model to represent multifractal characteristics in designated short time scales.