A comparative study on simulation models for self-similar traffic

Abstract

The explosive growth in Internet traffic has stressed the importance of the study of stochastic models enabling to represent the self-similar nature of this type of traffic. This paper presents a comparison of some models and algorithms that may be used to simulate self-similar traffic, namely the M/G/∞ process, the aggregation of ON/OFF reward renewal processes, Fractional Autoregressive Integrated Moving Averages (F-ARIMA) and Fractional Gaussian Noise (FGN). The comparison is based on the complexity of the implemented algorithms, the processing times, the accuracy of the estimation of the Hurst parameter H and possible limitations and advantages of the underlying models. A discussion of possible correction techniques for the estimation of H using the variance-time plot is also presented. © Springer-Verlag Berlin Heidelberg 2003.