Chaotic Maps As Models of Packet Traffic

Recent packet traffic measurement studies have indicated the presence of significant statistical features which are more characteristic of fractal processes than conventional stochastic processes. We demonstrate the feasibility of modeling these features efficiently using deterministic chaotic maps. We pr esent results fr om several maps to i llustrate the traf fic characteristics that can be modeled, including a two parameter nonlinear map that captures several fractal properties. We further outline a performance analysis method based on chaotic maps that can be used to assess the t raffic significance of fractal propert ies. It is our conclusion that while there are considerable analytical dif ficulties, chaotic maps may allow accurate, yet concise, models of packet traffic, with some potential for transient and steady state analysis.