Analysis and simulation of WAN traffic by self-similar traffic model with OMNET

An OMNET package with an implementation of a useful modification of a well-known self-similar traffic model is presented. The package is designed for the analysis of real WAN traffic traces, for the identification of model parameters and for the generation of 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. The performance of the package was tested with traffic traces from CAIDA and Internet Traffic Archive collections. These experiments revealed that the probability distribution of the modern WAN traffic has multimodal “saw-like” form. The reason for that interesting phenomenon is yet to be discovered.