Flow-Decomposition for Performance Prediction of GPS Mechanism under Homogeneous Self-Similar Traffic

Abstract

The multifractal-like behaviour (i.e., self-similarity) of network traffic is highly detrimental to the performance of scheduling mechanisms and user-perceived Quality-of-Service. In order to accurately evaluate and predict the effects of self-similar traffic on the performance of the Generalized Processor Sharing (GPS) principle, this paper proposes an original and cost-efficient flow-decomposition approach to modelling the GPS system under homogeneous self-similar traffic (i.e., with the identical Hurst parameters). Based on the comprehensive analysis of the excess service sharing behavior of traffic flows, we decompose the GPS system into a group of single-server single-queue systems and investigate the queue length distributions for individual traffic flows. Extensive simulation experiments are conducted to validate the accuracy of the analytical results for predicting the performance of GPS systems.