Optimal Schedulable Region Analysis of FCFS Network Model Supports End-to-end Delay Requirement

Abstract

The modern high-speed integrated service networks have provided multimedia applications. There are diverse traffic characteristics and different quality of service requirements. A scheduling discipline should permit easy admission control. A switch controller should be able to make decision. For a given current set of connections and the description for a new connection, whether it is possible to meet the new connection's performance bounds without degrading the performance of exiting connections. By analysis the schedulable region [1], we can easily do admission control and measure the efficiency of a scheduling discipline. A network structure using FCFS scheduling with homogenous source model has been proposed in [2] and the number of the connection vs. delay bound is solved. But in the real world, the sources are heterogeneous. For example, we can lower the quality of MPEG picture of video conference and lower the data transferred to the network and require same delay bound. In this paper, we extend the previous work of [2] and derived the schedulable region of N classes QOS service. The connections of same class have same (! , ") traffic parameters. By this traffic model, we can fit the requirement of MPEG data and get the same delay. We derived a formula for delay bound requirement and the number of connections of each class. We found that with the delay bound as the QOS metric, the schedule discipline of FCFS meets the optimal allocation policy (OPT)[4]. We also developed an evaluation strategy RG (Relative Gain)with weighting ! to analyze the network utilization. The numerical results of tandem network model, show the proposed FCFS network structure is optimal.