Queueing Property for Different Type of Self-Similar Traffics

Queueing behavior with long-range dependence input is important to determine the queue resource on the intermediate bottleneck nodes. The end-to-end performance especially depends on the provisioning of queue resources and link bandwidth. The purpose of this research is to explore queueing properties for self-similar traffic using the network simulator. In various ways to generate the self-similar traffic such as the transferring traffic of files with self-similar distribution, we adopt the on/off traffic over TCP on the ns-2 network simulator. The simulation traffic obeys the Pareto distribution and observes queue size and other performance parameters. We generated the two different types of traffics; the small amount of traffic and that of large amount, and extracted the following results. Firstly, for the small amount of traffics, queue size consumption of the maximum queue size suddenly appears in terms of small increase of traffic flows and does not show the differences for the variations of shape parameter (alpha) under restricted maximum queue size. On the other hand, the case of the middle alpha=1.65 consumes the queue resource under the unrestricted queue sizes. Secondly, for the large amount of traffics, the queue size consumptions do not depend on the variation of alphas and fluctuate with cyclic manner.