Design of plug-in schedulers for a GridRPC environment

Grid middleware performance is typically determined by the resources that are chosen. Different classes of applications need different metrics to define a meaningful notion of performance. Such metrics include application workload, execution time estimation, disk or memory space availability, etc. In the past, few environments have allowed schedulers to be tuned for specific application scenarios. Within the Diet (Distributed Interactive Engineering Toolbox) project, we developed an API that allows the resource broker to be tuned for specific application classes. In a seamless way, generic or application-dependent performance measures can be used within the hierarchy of resource brokers.