Issue No. 04 - April (1994 vol. 5)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/71.273047
<p>A main objective of scheduling independent jobs composed of multiple sequential tasks in shared-memory and distributed-memory multiprocessor computer systems is theassignment of these tasks to processors in a manner that ensures efficient operation ofthe system. Achieving this objective requires the analysis of a fundamental tradeoffbetween maximizing parallel execution, suggesting that the tasks of a job be spreadacross all system processors, and minimizing synchronization and communicationoverheads, suggesting that the job's tasks be executed on a single processor. Theauthors consider a class of scheduling policies that represent the essential aspects ofthis processor allocation tradeoff, and model the system as a distributed fork-joinqueueing system. They derive an approximation for the expected job response time,which includes the important effects of various parallel processing overheads (such astask synchronization and communication) induced by the processor allocation policy.</p>
Index Termsparallel architectures; shared memory systems; distributed memory systems; queueingtheory; scheduling; synchronisation; processor allocation; multiprogrammed;distributed-memory; parallel processing systems; shared-memory; scheduling;independent jobs; multiple sequential tasks; job response time; distributed fork-joinqueueing system; processor allocation policy; bulk arrival queues; communicationoverhead; distributed memory; fork-join queues; modeling and analysis;multiprogramming; operating systems; parallel processing; synchronization delay
M. Squillante, S. Tripathi and S. Setia, "Analysis of Processor Allocation in Multiprogrammed, Distributed-Memory Parallel Processing Systems," in IEEE Transactions on Parallel & Distributed Systems, vol. 5, no. , pp. 401-420, 1994.