Improving the performance of fault-aware scheduling policies for desktop grids (be lazy, be cool)

Desktop Grids have proved to be a suitable platform for the execution of Bag-of-Tasks applications but, being characterized by a high resource volatility, require the availability of scheduling techniques able to effectively deal with resource failures and/or unplanned periods of unavailability. Fault-aware scheduling, proposed in [2], can be considered a promising approach, yielding to both performance improvements for Bag-of-Task-Applications and increased utilization for Desktop Grids. The best fault-aware scheduling strategy available at the moment uses on-line scheduling, that is it starts a task as soon as a machine becomes available. In this paper we present a machine selection policy based on the idea that sometimes is better to wait for another machine rather than greedily exploit an immediately available one. An extensive simulation study, carried on for a variety of realistic Desktop Grid configurations and Bag-of-Task workloads, has revealed that the new scheduling strategy further improve application performance and machine utilization with respect to the best fault-aware scheduling strategy among those proposed in [2].