Cost-Efficient Scheduling of Workflow Applications with Deadline Constraint on IaaS Clouds

Abstract

Nowadays, the vast majority of workflow applications are deploying on clouds for fast execution. Meanwhile, the market-oriented and price-driven characteristics of cloud computing make cost become a factor that cannot be ignored and challenge traditional workflow scheduling algorithms which focus only on the optimization of finish time (a.k.a makespan). A general way to consider cost and makespan at the same time is to model the problem as a constrained optimization problem. In this paper, we study the deadline-constrained and cost-minimization workflow scheduling problem, and propose the cost-efficient scheduling with deadline constraint (CESDC) algorithm. CESDC is a typical list scheduling algorithm, and contains three scheduling phases: deadline distribution, task prioritization and service selection. CESDC firstly distributes deadline to tasks by their levels and workloads, then prioritizes tasks according to a modified upward rank, and finally assigns services to tasks which meets the sub-deadline and minimizes the cost. Experiment results demonstrate that CESDC performs better in terms of success ratio and cost than those of several state-of-the-art approaches.