Self-similar traffic modeling in modern communication systems

Abstract

The traffic is modeled as a stochastic process representing the demand that users of a communications network impose on the resources of the network. Originally it was considered that the times between arrivals of the demands of the users were independent of each other, as well as the same quantity of the demand (times between calls and duration of the calls, times between arrival of packages and length of the packages, times between requests for connections and length of sessions, etc.). Later, it was necessary to include the effect of the existing correlation between these variables, for which more elaborate models were developed in which the correlation declined exponentially with time. However, it has recently been shown that in modern communications networks, the correlation between these variables does not decay as rapidly and can persist across many time scales. This phenomenon, which significantly affects the performance of communications networks, can be adequately represented by fractal or autosimilar and multifractal traffic models. This article presents the basal concepts that are involved in the study of these models, in order to facilitate the reader introduction to the specialized literature on the subject.