Self-adaptive techniques for the load trend evaluation of internal system resources

Modern distributed systems that have to avoid performance degradation and system overload require several runtime management decisions for load balancing and load sharing, overload and admission control,job dispatching and request redirection. As the external workload and the internal resource behavior of themodern system is highly complex and variable, selfadaptive techniques require a stable vision of the system behavior. In this paper we propose a trend modelthat guarantees a robust interpretation for load-awaredecision algorithms. Various experimental results in aWeb cluster demonstrate that the proposed models andalgorithms guarantee better stability of the load and areduction of the response time experienced by the users.