Performance analysis of hybrid polling schemes with multiple classes of self-similar and long-range dependent traffic input

Queuing delay have significant impact on the performance of network applications. To meet distinct delay requirements of multi-class end-user traffic (e.g., voice, streaming video, interactive gaming and email etc.),various queuing and scheduling schemes have been proposed. These schemes are analogous to a polling mechanism in which multiple traffic queues are concurrently handled by a single scheduler. However, researchers were unable to analyze this synergy between the conventional queuing-cum-scheduling and pollingmodels. Moreover, research on analyzing polling models assumed traditional Poisson traffic distribution whichis unable to capture self-similar and long-range dependent (LRD) characteristics and hence yield misleadingresults. Furthermore, published work related to self-similar traffic modeling is mainly based on conventionalqueuing-cum-scheduling which are simple approximations. The objective of this work is to analyze differentcombinations of conventional queuing and polling models that can satisfy distinct requirements of various kindsof applications in heterogeneous networks. In this paper, we exploit the synergy between traditional queuingcum-scheduling and polling models. We analyze different combinations of queuing and polling mechanismswith realistic traffic distributions i.e., self-similar and LRD. First, we develop an analytical framework for G/M/1 queuing system which contemplates multiple classes of self-similar and LRD traffic as input. We formulate the Markov chain for G/M/1 queuing system and extract closed-form expressions of queuing delay forcorresponding traffic classes. First, we analyze a combination of limited service polling model with non-preemptive priority queuing. We also analyze different combinations of polling models (i.e., exhaustive, gated and limited service). We validate the performance of the proposed analytical framework through simulations.Simulation results suggest that synergy of polling and scheduling dangle promising results.