Modeling and simulation of self-similar data traffic

Abstract

Traditional analysis of network traffic generally neglects the existence of self-similarity. Many queue models such as Poisson process often underestimate the burstiness of traffic. Recent studies have revealed that self-similarity has a profound impact on network performance. In this paper, following the discussion of Poisson model and several commonly used self-similar models, an FTP traffic generation model is proposed and simulated by NS-2, The results demonstrated that the superposition of many Pareto-type sources can generate statistical self-similar data traffic and the computing cost is acceptable.