Performance Results For Analytic Models Of Traffic In Telecommunication Systems, Based On Multiple ON-OFF Sources With Self-Similar Behavior

摘要

A class of stochastic point processes, called N-Burst models, is introduced that describe traffic in telecommunication systems as the superposition of up to N individual bursts. By using so called truncated Power-Tail distributions, exact results for performance parameters in analytic queueing models with self-similar arrival processes are derived and discussed in the second part of the paper. The observation of the mean Cell Delay and the Buffer Overflow Probability reveals distinct critical utilization values (blow-up points) at which both performance parameters radically increase when self-similar properties of the cell stream are involved. The cell-rate during the individural bursts has major impact: In the limits for very large or very low cell-rates, approximate models deliver adequate results. In real life however, network componeuts are expected to operate in the intermediate region, where the blow-up points are contained. Thus, exact modeling on cell-level, as done in the N-Burst model, is shown to be essential.