Abstract
The tradtional traffic models are not suitable for high speed network traffic with longrange dependence, self-similarity and power law distribution of interarrival times. In this study, to coincide with actual data, the variable arrival rate of high speed networks is modeled by fractal renewal process (FRP). Then a multiscale (multiarrival rate) representation is introduced for FRP. Based on this multiscale framework, the variable arrival rate can be decomposed into a mixture of prototype Poisson processes. In turn, this representaion leads to efficient new algorithm for the analysis, the synthesis and the estimation of high speed network traffic. Based on multiscale representation of FRP (MRPRP), the expectation maximization (EM) algorithm is employed for parameter estimation of the MRFRP-based traffic model from actual interarrival time measurements at Bellcore. Then the results will be employed for performance evaluations in high speed communication networks.