Processor scheduling policies for distributed memory systems canbe divided into space sharing or time sharing policies. In spacesharing, the set of processors in the system is partitioned and eachpartition is assigned for the exclusive use of a job. In time sharingpolicies, on the other hand, none of the processors is given exclusivelyto jobs; instead, several jobs share the processors (for example, in around robin fashion). There are advantages and disadvantages associatedwith each type of policy. Typically, space sharing policies are good atlow to moderate system loads and when job parallelism does not varymuch. However, at high system loads and widely varying job parallelism,time sharing policies provide a better performance. We propose a newpolicy that is based on a hierarchical organization that incorporatesthe merits of these two types of policies. The new policy is a hybridpolicy that uses both space sharing as well as time sharing to achievebetter performance. We demonstrate that, at most system loads ofinterest, the proposed policy outperforms both space sharing and timesharing policies by a wide margin
展开▼