The interaction between memory allocation and adaptive partitioning in message-passing multicomputers

Abstract

Most studies on adaptive partitioning policies for scheduling parallel jobs on distributed memory parallel computers ignore the constraints imposed by the memory requirements of the jobs. In this paper, we first show that these constraints can have a negative impact on the performance of adaptive partitioning policies. We then evaluate the performance of adaptive partitioning in a system where these minimum processor constraints are eased due to the provision of support for virtual memory. Our primary conclusion is that any performance benefits resulting from the easing of minimum processor constraints imposed by the memory requirements of jobs will be negated by the overhead due to paging.