A Proposal of Job-Worker Assignment Algorithm Considering CPU Core Utilization for User-PC Computing System

Abstract

As a low-cost high-performance master-worker-model-based computing platform for group members, we have studied the User-PC computing system (UPC). The UPC master assigns queuing jobs from users to UPC workers that use idling computing resources of members’ personal computers (PCs). In this paper, we propose a job-worker assignment algorithm to minimize the makespan, considering the number of job threads and the number of CPU cores. For evaluation of the algorithm, we conducted experiments running 72 jobs on the UPC system with six workers that have various numbers of threads and CPU cores. The schedules by the algorithm could significantly reduce the makespan compared to other algorithms.