A genetic algorithm approach for utility management system workflow scheduling

Abstract

In this paper, we present a scheduler for distributing workflows in Utility Management System (UMS). The system executes a large number of workflows, which have very high resource requirements. The workflows have different computational requirements and thus the optimization of resource utilization must be performed in a way that is different from the standard approach of scheduling workflows. We developed a strategy for allocating workflows, which is based on a genetic algorithm. The proposed architecture executes a scheduling algorithm by using a feedback from the execution monitor. We also report on an experimental study, which shows that a significant improvement of overall execution time can be achieved by using the genetic algorithm. The algorithm is used for designing effective Grid schedulers that optimize makespan. The study further shows that the overall system (UMS) performance is significantly improved; this finding indicates that there can be reduction in hardware investment.