Slack-based advance reservation for Grid jobs

Advance reservations are usually used to guarantee the QoS of Grid applications by reserving a particular resource capability over a defined time interval on local resources. AR generally results in lower utilization of local resources and longer waiting time of non-AR jobs by interrupting processing time- and/or space- frames. This paper presents a slack-based AR mechanism, in which every AR request has a slack time to determine how long it may wait to run actually after its required start time. Different from existing studies in which slack time is simply used to relax the start time of an AR request from a rigid start time ST to a time span [ST, ST+slack], the mechanism in this paper also allows to postpone the execution of accepted reservations within their slack limits in order to accommodate more AR or local jobs. Experimental results show that it is helpful to increase the acceptance ratio and scheduling flexibility of AR jobs, and meanwhile improves the utilization of local system, reduces the disadvantageous impact of AR jobs on local jobs´ waiting time and slowdown.