All request shapes non-contiguous submesh allocation strategy for 2D mesh multicomputer

Abstract

Contiguous sub-mesh allocation usually suffers from the degrading effects of processor fragmentation as it requires that the processors that are allocated to a job request should be contiguous and have the same topology as the network connecting these processors. In non-contiguous allocation, the size and shape constraint affects on the system performance in terms of parameters such as job turnaround time and system utilization. Most recent non-contiguous allocation strategies suggested for 2D mesh multiconputers change the orientation of the job request when allocation fails for the original job request, and hence the system performance is improved. To generalize this restricted rotation, we propose in this paper, a new noncontiguous allocation strategy for 2D mesh multicomputers, referred to as All Request Shapes Greedy Available Busy List (ARSGABL for short), which considers all possible request shapes when attempting allocation for a job request. ARSGABL depends on the non-contiguous allocation strategy proposed in [20] for selecting an allocation sub-mesh. Results from extensive simulations under a variety of system loads and different job sizes confirm that the ARSGABL strategy improves system performance in terms of the average turnaround times of jobs.