Using Slack Reservation Strategy to Improve User QoS in Computational Grids

Advance reservation is an effective technique to realize end-to-end quality of service (QoS) guarantees for grid applications. However, excessive advance reservation will result in low resource utilization and high rejection rate on receiving new requests, which in turn reduces user’s QoS satisfaction. In this paper, a slack reservation strategy is proposed, with aiming to mitigate the negative effects brought about by conventional reservation service. The strategy is based on the observation that applications tend to overestimate reservation deadline to ensure their completion. It allows accepting new reservation requests that overlapping with existing ones under certain conditions. Both the benefits and risks of the proposed policy are presented theoretically. Experimental results show that the strategy can bring about remarkably higher resource utilization and lower rejection rate when the system is in presence of heavy reservation workload.