基于分布式多引擎架构的网格工作流管理系统  

A Grid Workflow Management System Based on Distributed Multi-Engine Framework

作　　者：赵钢[1] 

机构地区：[1]西安航空职业技术学院,西安710089

出　　处：《西南大学学报（自然科学版）》2012年第11期100-107,共8页Journal of Southwest University(Natural Science Edition)

基　　金：陕西教育厅基金资助项目(11JK0487)

摘　　要：主要研究面向服务的网格工作流管理系统,扩展了服务流程定义语言BPEL,使其支持将物理服务和虚拟服务作为原子服务参与构建工作流服务;提出了分布式多引擎系统架构;基于M/M/m排队模型,设计了负载均衡调度算法,能够在各引擎之间进行作业优化调度.实验结果表明,与集中式单引擎系统相比,分布式多引擎系统减少了作业平均响应时间,增强了系统负载能力和可靠性,提高了作业吞吐率和执行成功率,同时具备良好的可扩展性.Currently most grid workflow systems are built in a centralized single engine with low workload capacity, single point of failure and poor scalable performance. This paper describes ServiceFlow, a grid service-oriented workflow management system, which extends BPEL to support Web service, WSRF serv- ice and Virtual service participating in grid service composition, and adopts a distributed multi-engine framework, in which, based on M/M/m queuing model, load-balancing scheduling algorithms are pro- posed to dispatch jobs among engines. The results of performance evaluation show that ServiceFlow can a chieve shorter mean job response time, higher throughput and better dynamic scalability compared to the centralized single-engine workflow system.

关 键 词：网格工作流 M M m排队模型 负载均衡 作业调度 

分 类 号：TP393.028[自动化与计算机技术—计算机应用技术]