A distributed workflow management model for grid middleware

Abstract

Workflow management enables grid service composition and user collaboration in grid middleware, it includes the workflow of grid service invoking and data exchanging among services. Because of the large requirements of grid service composition, it's necessary for the workflow management to be scalable in many grid application systems. This paper proposes a distributed workflow management model for grid middleware, it is comprised of three modules including workflow definition tool (WDT), workflow balancer (WB), workflow execution engine (WEE) and defines the interfaces between workflow management and other grid components. Firstly, WDT composites grid services into workflow service, secondly WB dispatches workflow services and workflow job requests, thirdly WEE is up to the execution of workflow jobs. Furthermore, it describes the interfaces between workflow management and related grid components, such as job management, information service and grid portal. Also it divides the interfaces into WEE management, workflow service management and workflow job management. With the implementation and experiments of this model for Chinese education and research, its support platform middleware, its usability is verified through the usage of workflow tools in image processing scenario, its scalability and load balancing ability are proved in workflow scheduling test.

Keywords:

• grid computing
• grid middleware
• workflow service
• distributed workflow management

Acknowledgements

This paper is supported by National Natural Science Foundation of China (60503066), National Grand Fundamental Research 973 Program of China (2006CB303000), National High-Tech Research and Development Plan of China (2006AA01Z311), China Education and Research Grid Program (ChinaGrid), Programfor New Century Excellent Talents in University (NCET-05-0187), Program for Beijing New Stars in Science and Technology (200411A).

We would like to acknowledgment Ke Hao and the members of CGSP group including Bin Cai, Haijun Cao for their supports on this paper.