Comparison of predictive techniques in cluster-based network servers with resource allocation

Abstract

Predictive algorithms play a crucial role in system management by alerting it to future events that may degrade the global performance of the system. Our paper proposes a predictive algorithm for a Web switch that manages a cluster of Web servers. Several throughput predictive techniques have been studied to balance the workload dynamically according to a predetermined quality of service for different types of users, and finally we have obtained the best that fits in our algorithm. In order to reduce the checking time, estimations are only computed in a variable slot scheduling.