云环境下面向负载均衡的数据密集型工作流的数据约简策略  

Data reduced strategy for load-balanced data-intensive workflow in clouds

作　　者：胡志刚[1] 李佳 郑美光[1] Hu Zhigang;Li Jia;Zheng Meiguang(College of Software Engineering,Central South University,Changsha 410075,China)

机构地区：[1]中南大学软件工程学院

出　　处：《计算机应用研究》2019年第8期2410-2414,2420,共6页Application Research of Computers

基　　金：国家自然科学基金资助项目(61602525,61572525)

摘　　要：如何对数据密集型工作流应用进行高效合理地调度成为云计算领域亟待解决的关键问题之一。针对此问题,构造数据密集型工作流的有向超图模型,提出了数据支持能力概念,通过基于数据支持能力的合并操作对模型进行约简。最后优化超图多层剖分算法,提出数据约简的数据密集型工作流调度策略HEFT-P。研究结果表明,HEFT-P相比典型的工作流调度策略HEFT、CPOP、MCP,能够很好地对数据密集型工作流进行约简优化,获得较少的调度时间。How to schedule data-intensive workflow efficiently and reasonable has become one of the key issue in cloud computing. To address this issue, first, this paper built a directed hypergraph model for data-intensive workflow. And it proposed a concept data supportive ability to help the presentation of data- intensive workflow application and provided the merge operation details considering the data supportive ability. By optimizing the hypergraph multi-level partitioning algorithm, it proposed a data reduced scheduling policy HEFT-P for data-intensive workflow. Through simulation, it compared the classical HEFT, CPOP and MCP scheduling policies with HEFT-P. The results indicate that HEFT-P can obtain reduced data scheduling and reduce the makespan of executing data-intensive workflows.

关 键 词：数据密集型工作流 有向超图 数据约简调度 云计算 负载均衡 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]