Dynamic scheduling for jobs in the grid environment

In the GRID computing environment, a job is processed by a cluster of computing devices called a Virtual Organization, or a VO. The jobs are scheduled to VOs on the first-come-first-serve basis. Though the scheduling policy maintains the fairness of the job processing order, in some cases it might result in the under-utilization problem. This problem results from that some computing devices are idle until their assigned tasks are ready to be processed. Moreover the jobs are issued continuously and the number of jobs is unknown, thus a dynamic scheduling method is needed for the GRID environment. In our research, we proposed a dynamic scheduling technique based on the greedy methods. A user-job is decomposed into tasks which are assigned to the available processing elements in the GRID according to their precedence and priority. Our experimental results showed that our mechanism has better performance than the first-come-first-serve policy, in terms of a job turnaround time and the GRID's utilization rate.