Deadline Constrained Load Balancing Level Based Workflow Scheduling for Cost Optimization

Abstract

The advent of Cloud computing has provided a<br><br>promising methodology for usage of distributed resources for<br><br>complex scientific workflow applications. Due to the unique<br><br>features of cloud technology, such as the pay-as-you-go pricing<br><br>model and scaling, efficient workflow scheduling is a critical<br><br>research topic. While most workflow scheduling algorithms are<br><br>proposed to minimize the overall execution time, cost-driven<br><br>public cloud services have made cost minimization an<br><br>emerging and critical issue. Therefore, the objective of this<br><br>work is to solve the cost optimization problem for scheduling<br><br>workflows on a commercial Cloud while considering a userdefined<br><br>deadline constraint. In this paper, a heuristic<br><br>algorithm for scheduling deadline-constrained workflows is<br><br>presented - Deadline Constrained Level Based (DCLB)- which<br><br>uses Level Load Balancing to refine deadline distribution as<br><br>well as attaining lower communication cost in order to reach<br><br>the algorithm's goals. Experimental results demonstrate that<br><br>DCLB compared to existing algorithms, achieves higher cost<br><br>efficiencies when workflow deadline is met.