One of the most important issues in cluster computing systems is the efficient use of resources to increase the performance of systems and hence decrease their response times. These objectives can best be pursued by job schedulers in cluster computing systems. However, existing schedulers in cluster computing systems do not use resources efficiently. This paper proposes a new method for efficient allocation of submitted jobs to resources. Jobs consist of threads that are arranged in a two-dimensional matrix. Using the horizontal scanning of this matrix, threads of different jobs are allocated to different processors, preventing resources becoming idle. In previous method of scheduling, the resources are allocated to total threads of a job synchronization but in proposed method, the resources are allocated to threads of various jobs. In new method if there aren’t available resources enough for a job, threads of different jobs can run thus waste of resources are minimum. Simulation results of our proposed scheduling method show quicker cluster system response time than FCFS and Backfilling scheduling methods.
展开▼