Analysis of Queueing Property for Self-Similar Traffic

The scale-invariant burstiness or self-similarity has been found in real network. Relation between self- similarity and network parameter is mainly discussed in the context of application layer, and this self- similarity is caused by the file size of Web servers or the duration of user sessions. Traffic dynamics, however, are mainly generated by physical conditions such as the resource restrainment and the queue size on intermediate routers. The purpose of this research is to provide the basic concept of resource provisioning. In this paper, we have investigated queueing property on the bottleneck link for self-similar traffic using network simulator. We examined the distribution property of Pareto distribution on ns-2 simulator, then analyzing the simulated results, following properties were extracted. Firstly, the traffics with self-similar property are transmitted consuming the queue resources less on the bottleneck link. Secondly, queueing dynamics tend to exceed the linear manner for increase of packet flows. Thirdly, bandwidth expansion improve the consumption of the bottleneck queue while small delay does not affect to consume the queue. Finally, the effective throughput performance gain is measured under the restricted resource conditions.