The explicit analysis of polling systems combined with non-preemptive priority scheduling under self-similar traffic input

Abstract

The polling model resembles with a system of multiple queues targeted by the single server in cyclic order. Polling model has effective impact over performance investigation of network related applications. To date, the polling models have been extensively studied and analyzed only with Poisson traffic distributions. Also, in terms of providing guaranteed QoS to the end-user of multiservice Internet, a variety of scheduling schemes have been introduced in the market such as Priority Queuing (PQ), Custom Queuing (CQ), Weighted Fair Queuing (WFQ), Class Based Weighted Fair Queuing (CBWFQ) and Low Latency Queuing (LLQ) etc. All these schemes somehow represent some sort of polling mechanism in which a single server serves multiple queues in a cyclic fashion according to a specific scheduling logic. To date, the researchers have been analyzing the polling models and scheduling schemes separately. The novelty, this study brings in the area of polling models is twofold. First, we analyze different kinds of polling schemes (exhaustive, gated and combination of both) with realistic traffic distributions i.e. self-similar and long range dependent. Second, in our recent work, we have found that the combination of traditional polling schemes (exhaustive, gated and limited service) with common scheduling schemes such as priority can offer a differential treatment in terms of providing the required QoS to different kinds of applications found in modern 4G wireless networks.