面向Hadoop平台的工作流建模与实施  

Hadoop Oriented Data-intensive Workflow Modeling and Enactment

作　　者：席永涛[1] 李奇原[1] 刘杰[1] XI Yong-tao,LI Qi-yuan,LIU Jie（Technology Center of Software Engineering,Institute of Software,The Chinese Academy of Science,Beijing 100190,China）

机构地区：[1]中国科学院软件研究所软件工程技术中心,北京100190

出　　处：《电脑知识与技术》2012年第7期4635-4640,共6页Computer Knowledge and Technology

摘　　要：大数据处理和分析技术能够指导企业进行快速、正确的决策,将在企业业务流程中发挥重要的作用。为了支持企业开发数据密集型应用,并将其集成到已有的业务流程中,需要基于Hadoop平台的数据密集型工作流系统,其中的核心技术之一是面向Ha doop的工作流建模。该文基于BPEL语言定义了Hadoop工作流建模语言,使得该工作流建模语言支持丰富的流程控制语义,并通过模型转化为BPEL流程,可以运行于成熟的BPEL引擎中,实现可靠高效的工作流调度。该文还详述了Hadoop工作流到BPEL在语义模型、部署模型、执行模型等三个层次的转换。最后,通过一个日志分析工作流案例展示了方法的可用性。Big data processing and analysis is playing an important role in the enterprise business process management by directing lot of en terprises to carry out rapid and correct decision.To satisfy the needs of enterprises to develop these data-intensive applications and integrate these into existing business processes,the data-intensive workflow systems based on Hadoop are required to develop,in which one of main techniques is workflow modeling and enactment for Hadoop.In this paper,a Hadoop workflow modeling language is defined based on WS-BPEL,which makes the workflow modeling language supporting a rich of flow control semantics,and by transforming it into a BPEL process,the workflow model can be easily run in a mature BPEL engine with reliable and efficient workflow scheduling.The transforma tion from Hadoop Workflow to BPEL process at three levels of the semantics,deployment and execution,is also presented.Finally,an ex ample of the log analysis workflow is given to demonstrate the availability of our work.

关 键 词：业务流程管理 数据密集型计算 工作流系统 Hadoop BPEL 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论;自动化与计算机技术—计算机科学与技术]