一种任务合并机制下的云工作流多阶段调度方法  

MULTI-STAGE SCHEDULING METHOD OF CLOUD WORKFLOW WITH TASKS MERGING

作　　者：黄志刚[1] 刘峰[2] Huang Zhigang;Liu Feng(Hunan Vocational College of Science and Technology,Changsha 410004,Hunan,China;School of Computer,Changsha University of Science and Technology,Changsha 410082,Hunan,China)

机构地区：[1]湖南科技职业学院,湖南长沙410004 [2]长沙理工大学计算机学院,湖南长沙410082

出　　处：《计算机应用与软件》2019年第12期19-28,共10页Computer Applications and Software

基　　金：国家自然科学基金项目(51508041)

摘　　要：云环境中的工作流调度问题不同于传统分布式计算环境,主要体现在按需资源提供机制与即付即用的资源定价模式。为了解决这一问题,提出一种动态的代价优化工作流调度算法。为了适应用户对完成工作流的截止时间限制,对工作流结构中的管道任务进行了合并,减少了管道任务部署于同一节点时的数据传输时间;通过循环控制机制中代价最低预调度、资源提供决策及正式调度三阶段方式,生成了满足截止时间的代价优化调度解。实验结果表明,与同类算法比较,当设置不同的截止时间因子改变约束程度时,该算法在截止时间约束满意度以及调度时间和调度代价的同步优化方面均优于对比算法。The workflow scheduling problem in the cloud environment is different from the traditional distributed computing environment,which is mainly reflected on the on-demand resource provision mechanism and pay-as-you-go resource pricing model.For solving this problem,we present a dynamic cost optimization workflow scheduling algorithm.In order to adapt to the user s deadline for completing the workflow,we merged the pipeline tasks in the workflow structure to reduce the data transmission time when the pipeline tasks were deployed on the same node.The algorithm generated the scheduling solution with optimal cost to meet the deadline through three stages of the least-cost pre-scheduling,resource provision decision-making and formal scheduling in the cyclic control mechanism.The experimental results show that,compared with other similar algorithms,the proposed algorithm has better deadline constraint satisfaction and better synchronization optimization performance on the scheduling time and the scheduling cost than the contrasted algorithms when different deadline factors are set to change the constraint degree.

关 键 词：云计算 代价优化 截止时间限制 任务调度 任务合并 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]