Skip to Main Content
There are many grid workflow systems appeared in the past few years. But most of them involved in many developers and need a long time to implement them. It is very difficult to small groups to build a grid workflow system with good performance. In this paper we review the architecture and technologies used by those grid workflow systems briefly, and present the state of the art of our on going research on grid workflow systems. And what is the most important thing in this paper is, we present a novel architecture for building a grid workflow system with fast implement speed and good performance. Using this architecture, we have implemented a grid workflow system: grid of USTB. The test bed show that the architecture works well with computing-sensed and data-sensed applications. The architecture is loose-coupling both vertically and horizontal. It is easy to implement, extend and modify. In order to got better robust and simplify the course of implementation, we integrate two great soft wares: Condor as the manager of underlying hardware, and OSworkflow as the local workflow engine. Moreover, we use Web service as the highest layer strategy.
Machine Learning and Cybernetics, 2008 International Conference on (Volume:2 )
Date of Conference: 12-15 July 2008