A ew Multifractal Traffic Model and Its Impacts on Buffer Queueing Performances

In this paper, we propose a new analytical expression for estimating byte loss probability at a single server queue with multi-scale traffic arrivals using Pareto distribution. In order to make the estimation procedure numerically tractable without losing the accuracy, we assume and demonstrate that an exponential model is adequate for representing the relation between mean square and variance of Pareto distributed traffic processes under different time scale aggregation. Extensive experimental tests validate the efficiency and accuracy of the proposed loss probability estimation approach and its superior performance for applications in network connection with respect to some wellknown approaches suggested in the literature