Contradictory relationship between Hurst parameter and queueing performance (extended version)

摘要: Long Range Dependent (LRD) network traffic does not behave like the generated by Poisson model or other Markovian models. From performance point of view, main difference is that LRD increases queueing delays due to its burstiness over many time scales. behavior has been observed in different types and sizes networks, for applications (e.g., WWW) aggregations. Since behaviour rare nor isolated, accurate characterization very important order predict allocate resources. The Hurst parameter commonly used quantify degree traffic. In this paper we investigate validity effectiveness parameter. To end, analyze UCLA Computer Science Department traces compute their parameters. Queueing simulation study impact determine if accurately describes such LRD. Our results show itself an predictor a given trace.