Performance analysis of mixed polling schemes with multiple classes of self-similar traffic input to build comprehensive SLAs

Abstract

Solitary polling technique is not a better choice to get rid of multiple queuing problems for getting enhanced performance against single server. Enhanced and reliable performance upon multi queued traffic can be achieved through the utilization of right selection of joint polling schemes. The complexity of polling design is directly proportional to its performance. Exhaustive polling scheme is outperformed by limited service polling policy but under average load of traffic both contain same mean transfer delay. Several polling methods are the part of literature but the question of their optimal utilization in joint manners against multiple queue buffers is still open. In this paper, we build a complete analytical framework for mixed exhaustive and limited service polling model based on G/M/1 queuing system. We build the Markov chain and extract the closed form expressions of packet delay for multiple classes of self-similar traffic. We further experimentally implement four queues model with different arrangements of mixed polling schemes to conclude the most optimal joint polling model for modern 4G wireless network through analyzing the co-relation between delay and Hurst parameter.