A Simulation Study of the Impact of Switching Systems on Self-similar Properties of Traffic

Recent research has shown that traffic in Ethernet and other networks tends to exhibit properties of self-similarity such as long-range dependence and a high degree of correlation between arrivals. This paper investigates the impact of the switching network on the self-similar properties of the traffic. This simulation study reveals that switching networks tend to reduce the self-similarity of highly self-similar traffic. This is because of the truncation of long bursts due to packet discards, and also because of aggregation of flows through concatenated rather than superposed bursts. On the other hand, switching systems have the opposite effect of increasing the self-similarity of input traffic that has no selfsimilar properties such as traffic with Poisson or uniformly random distributions. This paper also presents simulationbased evidence of the causes behind these phenomena.