Abstract: The high energy consumption brought by executing workflow tasks not only increases the economic cost of cloud resource providers, but reduces the realibility of cloud system. In order to meet the deadline and reduce the energy consumption of executing workflow, an energy-efficient workflow scheduling algorithm CWEES is presented. CWEES divides the energy-efficient optimization scheduling into three stages：the initial tasks mapping, the processors resource merging and the tasks slacking. The initial tasks mapping aims to get the intial tasks scheduling orders by using the down-up leveling ordering. The processors resource merging aims to reduce the number of used resources by reclaiming the slack time and merging the relatively inefficient processors. The tasks slacking aims to select the best available resource with appropriate voltage/frequency level for each task so that the total energy consumption is minimal while meeting its sub-deadline. Simulation experiments are constructed to evaluate CWEES’s performance by the random workflow tasks model. The results show that CWEES not only can obtain higher resource utilization, but can reduce the energy consumption of executing workflow, which could achieve the better trade-off between the execution efficiency and the energy consumption.

Key words: cloud computing, workflow scheduling, energy efficiency, task allocation, resource merging

摘要： 工作流任务执行时带来的高能耗不仅会增加云资源提供方的经济成本，而且会降低云系统的可靠性。为了满足截止时间的同时，降低工作流执行能耗，提出一种工作流能效调度算法CWEES。算法将能效优化调度划分为三个阶段：初始任务映射、处理器资源合并和任务松驰。初始任务映射旨在通过任务自底向上分级排序得到任务调度初始序列，处理器资源合并旨在通过重用松驰时间合并相对低效率的处理器，降低资源使用数量，任务松驰旨在为每个任务重新选择带有合适电压/频率等级的最优目标资源，在不违背任务顺序和截止时间约束前提下降低工作流执行总能耗。通过随机工作任务模型对算法的性能进行了仿真实验分析。结果表明，CWEES算法不仅资源利用率更高，而且可以在满足截止时间约束下降低工作流执行能耗，实现执行效率与能耗的均衡。

关键词: 云计算, 工作流调度, 能效, 任务分配, 资源合并