Análise de desempenho de redes de filas com tráfego modelado por distribuições de cauda pesada

Abstract

The changes that occurred in telecommunications networks in recent years also affected the statistical profile of the traffic of these networks. Markov models, previously used to analyze the performance of networks are considered to be inadequate, and thus there is the need to use models that take into account this traffic self-similarity. In this work we investigate the behavior of single queues and networks of queues, when the arrival process is modeled by a heavy tail distribution such as Pareto, lognormal or Weibull, and the service time is modeled by an exponential distribution or another heavy tail distribution, or when it is assumed constant. For single-queue systems, three different approaches to characterize the traffic were considered: fixed shape parameters and variable service time; fixed service time and variable form parameters; and variable number of traffic sources with fixed shape parameters and service times. For the networks of queues we considered systems without derivation/integration of traffic and intermediate systems in which there is such derivation/integration.