Modern global network traffic model for simulation system

Currently, the impact of traffic self-similarity on different aspects of telecommunications is still an important research topic. One of the important problems in this field is developing tractable traffic models and evaluating their ability to match closely real traffic behavior. The first purpose of this paper is to present OMNET++ realization of our modification of well-known self-similar traffic model «M/G/∞ Input». Modified model is able to capture several characteristics of real WAN traffic together, including Hurst parameter, probability distribution of traffic value and autocorrelation function (ACF) of traffic value. Another purpose of this paper is to make parameters of modified «M/G/∞ Input» model more tractable from the practical point of view. An OMNET package with an implementation of a useful modification of well-known self-similar traffic model is presented. The package is designed for analysis of real WAN traffic traces, identification of model parameters and generation of a synthetic traffic, which is close to real one by a number of characteristics simultaneously. These characteristics include traffic value probability distribution, normalized autocorrelation function and Hurst parameter. Performance of the package was tested with traffic traces from CAIDA and Internet Traffic Archive collections. An interesting phenomenon of multimodal «saw-like» form of the probability distribution of modern WAN traffic was reviled in these experiments.