Selflike traffic of multiservice communication networks: myths and reality

The article considers the possibilities of using self-similar process models for traffi c service systems analysis in multiservice communication networks. The queuing system analytical models are widely used when studying communication network traffi c. However, it is impossible to apply the Poisson fl ow models for the analytical study of modern multi-service packet-switched net-works. Requests in such networks are strongly correlated, and even non-Poisson request interval probability distributions lead to signifi cant errors when analyzing queues if their correlation prop-erties are not accounted for. Furthermore, self-similar processes that refl ect long-term correlation dependencies in the multiservice network traffi c cannot be used for a network traffi c analysis due to the complexity of the model.