Largest-job-first-scan-all scheduling policy for 2D mesh-connected systems

In general parallel computer systems, scheduling incoming jobs has been identified to be an important factor determining the overall performance, in addition to the allocation of processors to the scheduled jobs. The authors propose an efficient scheduling scheme for two-dimensional meshes. By employing the largest job first and scan-all policy along with the waiting time limit, the proposed scheme can alleviate the fragmentation problem. Contrary to the previous largest-job-first scheduling schemes, large jobs do not block small jobs in the scheme. As a result, the mean response time is significantly reduced as identified by comprehensive computer simulation.