On aggregate traffic generation with multifractal properties

This paper studies time scaling and queueing behavior of aggregate traffic that is asymptotically self-similar and also exhibits small time-scale irregularities introduced via a cascade construction. The main contribution of the paper is to provide some insight into how to choose an appropriate cascade construction in order to obtain desired scaling and queueing properties. It is found that the largest time-scale of mass redistribution has far more influence on queueing behavior than the number of stages in the cascade. Thus, appropriate queueing behavior can be achieved by using one or two stages of mass redistribution, perhaps derived from the actual network dynamics