Heuristic Scheduling Method with the Importance of Earlier Tasks for Deadline Constrained Workflows in Clouds

Cloud has gradually become a more popular platform for executing scientific workflow applications due to its unique features. However, how to effectively schedule scientific workflows in clouds to minimize execution cost under a deadline constraint remains a critical problem. And in this problem, deadline distribution strategy plays a key role. In this paper, a new deadline distribution strategy with level importance rate is proposed. Based on this, we propose a heuristic scheduling method, named LIRHM. LIRHM prioritizes tasks by their upward ranks and distributes the deadline to each task as its sub-deadline using the proposed deadline distribution strategy, then maps each task to the best VM instance which meets its sub-deadline and minimizes workflow execution cost. Compared to four methods with different deadline distribution strategies, experiments on five real-word scientific workflows show the effectiveness of LIRHM and the importance of earlier tasks on deadline distribution strategy. In addition, we analyze the performance difference of different deadline distribution strategies on each workflow.