Priority-Based Job Scheduling in Distributed Systems

Global computing systems like SETI@home tie together the unused CPU cycles, buffer space and secondary storage resources over the Internet for solving large scale computing problems like weather forecasting, and image processing that require high volume of computing power. In this paper we address issues that are critical to distributed scheduling environments such as job priorities, length of jobs, and resource heterogeneity. However, researchers have used metrics like resource availability at the new location, and response time of jobs in deciding upon the job transfer. Our load sharing algorithms use dynamic sender initiated approach to transfer a job. We implemented distributed algorithms using a centralized approach that improves average response time of jobs while considering their priorities. The job arrival process and the CPU service times are modeled using M/M/1 queuing model. We compared the performance of our algorithms with similar algorithms in the literature. We evaluated our algorithms using simulation and presented the results that show the effectiveness of our approach.