A Proposal of Dynamic Job Scheduling Algorithm Considering CPU Core Utilization for User-PC Computing System

Abstract

The User-PC computing system (UPC) has been studied as a very low cost distributed computing platform that leverages idling resources of users’ personal computers (PCs). Designed according to the master-worker model, the master PC accepts jobs from users and assigns them to available worker PCs. An efficient static job assignment algorithm considering CPU core utilization has already been devised for the UPC system. It finds an optimal assignment that minimizes the makespan in a static scenario, i.e. jobs are all available in the beginning and all worker PCs are idle. However, in reality, jobs dynamically join and leave the system and as such, some workers may be busy when new jobs arrive. In this paper, we propose a dynamic job scheduling algorithm for the UPC system by extending the static job assignment algorithm. It finds assigned workers to all queuing jobs either on new job arrival, or on job completion by a worker PC. For evaluation, we conducted experiments using six worker PCs and up to 42 jobs. Our algorithm could significantly reduce the makespan compared to other scheduling algorithms.