Simulation of the Harmful Consequences of Self-Similar Network Traffic

Abstract

For a long time, the area of network traffic research has lacked adequate traffic measurements. During the past years, large amounts of network traffic measurements have become available, collected in the Web and high-speed networks. Measurements of real traffic indicate that burstiness is present on a wide range of time scales and show evidence of self-similarity and long-range dependency. Many traditional models originating in conventional voice networks either assumed that the underlying time series are short-range dependent or do not focus on the various, interrelated network components affected by bursty traffic. In order to provide explanations for empirically observed phenomena, such as burstiness in terms of physical network parameters, we need to develop models that can allow for the details of the complex architecture of today's networks and can analyze the interrelated network parameters that ultimately determine the performance and operation of a network. In the paper, we investigate the harmful consequences of burstiness on various network components using a model implemented by a discrete event simulation methodology. The methodology based on a modified version of the M/Pareto model can measure network performance parameters, such as the momentary utilization of links, response time, and the queuing performance of interconnected switches and routers along the traffic paths. Our paper intends to narrow the gap between existing, well-known theoretical results and their applicability in everyday, practical network analysis and modeling. Our methodology can help network designers and engineers, the ultimate users of traffic modeling, understand the dynamic nature of network traffic and assist them to design, monitor, and control complex, high-speed networks in everyday practice.