PERFORMANCE EVALUATION OF NOVEL SCHEDULING ALGORITHMS FOR G/M/1 QUEUEING SYSTEM WITH MULTIPLE CLASSES OF SELF-SIMILAR TRAFFIC INPUT

The continued dramatic increase in different types of end-user traffic in the internet requires different performance characteristics in terms of guaranteed QOS. It has been shown through high quality studies that internet traffic exhibit self-similarity and burstiness over a large range of time scales. Unlike most existing studies that are generally contingent on simple rule of thumb such as Poisson models, this research introduces promising analytical performance models for multiple queue system with self-similar traffic input being scheduled by novel scheduling algorithms to support differentiated traffic classes with guaranteed QOS. Our models are substantiated on G/M/1 queueing system that takes into account multiple classes of traffic exhibiting self-similar and long-range dependent characteristics. We analyze the model on the basis of two different scheduling mechanisms and find different QoS parameters such as packet delays and packet loss rate for corresponding traffic classes. We also build a comprehensive discrete event simulator, which implements the traffic model and evaluate the QoS behavior of self-similar traffic under new proposed scheduling schemes.