A preemptive job scheduler based on a Backpropagation Neural Network

Abstract

This paper presents a preemptive job scheduler based on a 3-layer Backpropagation Neural Network (BPNN) and a greedy task alignment procedure. The BPNN estimates priority values of jobs based on the attributes of their subtasks and the given job selection criteria of the scheduler. The scheduler is formulated in such a way that, at each time interval, the most priority job will be selected from the job queue before the next job arrives. The selected job is only preempted by a new job if its priority is less than the new job and then the preempted job will be restarted when its priority reaches high. When a predefined threshold time is reached, the job queue is refreshed to eliminate the old and low priority jobs. The proposed satisfiability measure based on job validation test, BPNN convergence test and cost value assure the efficiency of the scheduler. The performed simulations show that the presented scheduler approach is an effective one for a preemptive job scheduling application.