【摘要】 针对多云环境下同时考虑任务执行故障和截止日期约束下的工作流应用调度问题，提出一种面向代价和可靠性的多目标快速非支配排序调度策略(RNSGA-Ⅱ),在满足工作流截止日期的前提下，降低了任务执行代价的同时提高系统执行的可靠性。该算法考虑受资源节点故障、网络拥塞等环境因素造成的任务执行失败问题，利用韦布尔分布对任务执行可靠性进行分析，并复制执行危险率较高的任务，将其分配至执行代价最低的服务实例上。通过考虑硬件故障、网络波动等干扰因素下的仿真实验结果表明，与现有的两种多目标优化调度策略相比，RNSGA-Ⅱ在执行代价上平均降低18.65%,在执行可靠性方面平均提高6.54%,同时在工作流截止日期满足率上具有更优秀的性能表现。更多还原

【Abstract】 We propose a multi-objective fast non-dominated sorting scheduling strategy(RNSGA-Ⅱ) for workflow application scheduling problems under the constraints of task execution failures and deadlines.The aim is to lower the cost of task execution while improving the reliability of system execution, with the prerequisite of meeting the workflow deadline in a cloudy environment.This algorithm considers the problem of task execution failure caused by environmental factors such as resource node failure and network congestion, analyzes the reliability of task execution using the Weibull distribution, and replicates and assigns high-risk tasks to the service instance with the lowest execution cost.Simulation experiments under the interference factors of hardware failures and network fluctuations show that, compared with two existing multi-objective optimization scheduling strategies, RNSGA-Ⅱ reduces average execution costs by 18.65% and improves average execution reliability by 6.54%,while having better performance in workflow deadline compliance rate.更多还原