The influence of load prediction methods on the quality of service of connections in the multiprocessor environment

This paper is devoted to the performance analysis of the multiprocessor teletraffic processing environment under control of different resource allocation algorithms. It is assumed that the quality of service of incoming traffic is measured as the average delay of connections imposed by processing. In order to minimize the average delay each incoming connection is directed to be processed on the server, the average load of which during processing of this connection would be minimal. Applied allocation algorithms differ in the method of prediction of servers average future load. It is shown on representative numerical examples that the choice of appropriate load predictor has significant impact on the quality of service delivered by the traffic processing system considered.