Self-similar traffic: a new model for next generation ATM switch

This paper gives a brief introduction to self-similar process and the estimation of Hurst parameter in the context of time series analysis. Different methods of self-similar traffic generation are compared based on a simulation study. We find that aggregating the streams of ON/OFF sources of Pareto distribution leads to self-similarity quickly as the sources become large. The quick traffic generation that approximates real traffic in various time scale is also discussed and its influence on the design, control and analysis of high speed cell-based ATM networks is learnt from the simulation.