FRAMEWORK FOR ADAPTIVE GRID COMPUTING SERVICE

This paper presents a framework for an adaptive Grid Computing Element service which can auto-scale its computational resources depending on the workload. To efficiently operate a Grid infrastructure the framework is designed to provide autonomic capabilities for the Grid service responsible for application execution. Self-management of the infrastructure is based on policy enforcement by a configuration management system. This capability is integrated with dynamic provisioning of the Grid site computational cluster on virtualized resources by a resource manager module using a feedback control loop. In order to efficiently allocate resources for the computing cluster we maximize the utility function for the job execution service. We have defined and experimented the dynamic allocation of resources so that the utility of the Grid Computing Element service is maximized.