A Novel Statistic-based Relaxed Grid Resource Reservation Strategy

Abstract

In grid environments, advance reservation mechanism is to provide reliable resource co-allocation and scheduling for applications. However, excessive advance reservation will bring about many negative effects for system's performance, such as lower resource utilization and higher rejection rate. To mitigate these negative effects of advance reservation, a relaxed reservation strategy is proposed, in which the reservation admission criteria is more relaxing than that of conventional reservation. The strategy is based on the facts that grid applications tend to overestimate reservation duration to ensure their completion. In the proposed relaxed reservation strategy, reservation manager might accept those reservations that are overlapping with existing ones. Experimental results show that the strategy can bring about remarkably higher resource utilization and lower rejection rate at the price of a slightly increasing of reservation violations.