PERFORMANCE EVALUATION OF A SCHEDULING ALGORITHM FOR PROVIDING QoS IN INTEGRATED SERVICES NETWORKS

Abstract

Integrated services networks face the challenge of managing several traffic classes at the same time. Service disciplines for integrated service network therefore need to be flexible i.e. able to provide different type of services in order to accommodate different traffic classes efficiently. When traffic flow pertaining to multiple traffic classes, each bearing different QoS requirements coexist in a network, resource management needs to be flexible enough to efficiently provide each flow with a different type of service according to its class at a feasible computation complexity. Service disciplines based on generalized processor sharing paradigm which schedule flows according to their weights lack flexibility needed to efficiently manage both classes at the same time. Fair queuing was invented to ensure that every flow gets it's fair share of total bandwidth. Fair queuing has proven to have nearly perfect isolation and fairness. Various schemes based on this idea have been developed. Though they provide good isolation and fairness most fair queuing schemes are too costly to implement as they take time to process each packet. Efficient fair queuing using Deficit Round Robin (DRR) reduces workload time per packet while maintaining fairness. Service disciplines based on dual class paradigm, which considers two traffic classes at the same time are flexible and efficient for integrated services networks. In this dissertation performance of Packet Timed Token Service Discipline (PTTSD) an innovative packet scheduling discipline, which approximates the dual class paradigm at a possible computational complexity has been compared with Deficit Round Robin (DRR) and DRR++ by simulation. Simulation results show that out of the three PTTSD is more efficient to provide QoS in integrated services networks. Simulation Program is written in Visual C++ on Pentium-IV PC having windows 98 operating system. Performance curves have been plotted in Microsoft Excel based on simulation data.