Online Algorithms for Single Machine Schedulers to Support Advance Reservations from Grid Jobs

Advance Reservations(AR) make it possible to guarantee the QoS of Grid applications by reserving a particular resource capability over a defined time interval on local resources. However, advance reservations will also cause the processing time horizon discontinuous and therefore reduce the utilization of local resources and lengthen the makespan (i.e., maximum completion time) of nonresumable normal jobs. Single machine scheduling is the basis of more complicated parallel machine scheduling. This study proposed a theoretic model, as well as four online scheduling algorithms, for local single machine schedulers to reduce the negative impact on the utilization of local resources and to shorten the makespan of non-AR jobs resulting from advance reservations for Grid jobs. The performances of the algorithms were investigated from both of the worst case and the average case viewpoints. Analytical results show that the worst case performance ratios of the algorithms against that of possible optimal algorithms are not less than 2. Experimental results for average cases suggest that the First Fit and the First Fit Decreasing algorithm are better choices for the local scheduler to allocate precedence-constrained and independent non-AR jobs respectively.