Extended flexible processor allocation strategy for mesh-connected systems using shape manipulations

Abstract

Current processor allocation techniques for mesh-connected parallel systems are restricted to rectangular submesh allocation strategies causing significant fragmentation problems. In this paper, we propose an EFPA (Extended Flexible Processor Allocation) strategy to reduce external fragmentation and job response time, simultaneously. EFPA manipulates the shape of the required submesh to a more generalized L-shaped submesh. When an incoming job requests a rectangular submesh, EFPA first tries to allocate the conventional rectangular submeshes as other strategies. If it fails, EFPA further tries to allocate more flexible and robust L-shaped submeshes instead of signaling the allocation failure. Thus, EFPA accommodates incoming job earlier than other strategies. All the shape manipulations to the L-shaped submeshes are transparent to the application programmers. Our simulations show that EFPA performs more efficiently than other strategies in terms of the external fragmentation and the job response time.