Modeling and Performance Analysis of Self-Similar Traffic Based on FBM

Self-similarity or long-range dependence (LRD) is a ubiquitous property of real traffic. Classical Poisson or Markov-based queue techniques are unsuitable for LRD traffic which creates the need for new analytical tools. Among the existing self-similar traffic models the fractional Brownian motion (FBM) process has been regarded as an attractive alternative to traditional modeling approaches, while it has some limitations. In this paper, we make a thorough investigation of traffic modeling and performance evaluation based on FBM. On the basis of the buffer overflow probability given by Norros, we derive the formulae of average queue length, queue length variance, average delay and jitter. The variation of these parameters is investigated through simulation, and some useful results are obtained. Finally, theoretical fractional traffic traces and real measured traffics traces are used to driven the OPNET to verify the accuracy and its scope of application of FBM model, and point out its inadequacy.