Task Scheduling Model

Abstract

To design and implement a task scheduling model which predicts a schedule for a new task set without actually running a task scheduling algorithm. Generating an optimal schedule of tasks for an application is critical for obtaining high performance in a heterogeneous computing environment and it is a hard problem. This work attempts to optimize on the scheduling time by designing a task scheduling model. The task scheduling algorithm used in this work is based on ACO, a swarm intelligence model. The prediction is done after the training phase of the model. The model is validated by comparing the predicted schedule with the actual schedule obtained by running the ACO scheduling algorithm on the new task set. The parameters used for comparison are waiting time of tasks, average processor utilization and the scheduling time. The predicted schedule is comparable to the actual schedule with respect to waiting time of tasks and average processor utilization. The scheduling time is significantly reduced and the reduction in the scheduling time increases with the increase in the task set size.