The Community for Technology Leaders
Green Image
<p>Managing computing resources in a hypercube entails two steps. First, a job must bechosen to execute from among those waiting (job scheduling). Next a particular subcubewithin the hypercube must be allocated to that job (processor allocation). Whereasprocessor allocation has been well studied, job scheduling has been largely neglected.The goal of this paper is to compare the roles of processor allocation and job schedulingin achieving good performance on hypercube computers. We show that job schedulinghas far more impact on performance than does processor allocation. We propose a newfamily of scheduling disciplines, called Scan, that have particular performanceadvantages. We show that performance problems that cannot be resolved throughcareful processor allocation can be solved by using Scan job-scheduling disciplines.Although the Scan disciplines carry far less overhead than is incurred by even thesimplest processor allocation strategies, they are far more able to improve performancethan even the most sophisticated strategies. Furthermore, when Scan disciplines areused, the abilities of sophisticated processor allocation strategies to further improveperformance are limited to negligible levels. Consequently, a simple O(n) allocationstrategy can be used in place of these complex strategies.</p>
Index Termsscheduling; resource allocation; hypercube networks; processor allocation; hypercubecomputers; hypercube; job scheduling; scheduling; Scan; performance problems

P. Krueger, V. Dixit-Radiya and T. Lai, "Job Scheduling is More Important than Processor Allocation for Hypercube Computers," in IEEE Transactions on Parallel & Distributed Systems, vol. 5, no. , pp. 488-497, 1994.
92 ms
(Ver 3.3 (11022016))