Modeling and Self-Similarity Analysis of Non-Poissonian Traffic Represented by Multimodal Non-Typical Pascal and Rice Distributions

Modern stage of development of information and communication networks requires solving of crucial tasks of network traffic statistical analysis and traffic simulation modeling. The predominance of non-Poisson traffic leads to the impossibility of analyzing multichannel communication systems by the methods of queuing theory that used to describe telephone networks. The last decade has paid much attention to research on traffic that has signs of self-similarity. The main purpose of this work is a statistical analysis of non-Poissonian traffic, represented by multimodal non-standard Pascal (negative binomial) and Rice distributions. As a result, a study of the self-similarity degree has been performed by the R/S analysis and the aggregation method. In addition, we propose to use EM-algorithm with an algorithm for determining an optimal number of clusters for an approximation of non-typical multimodal distributions.