Workflow Management Research Based On Virtual Service In Heterogeneous Grid System

As a new way to share resources, grid organically connects the distributed computing resources, storage resources and information resources. The research of grid interoperation aims at the difficulties that it is hard to share resources among heterogeneous grid systems. Heterogeneous grid workflow management integrates different grid system service resources, packages them as atomic services into a composite service, with larger particle size as well as more functions, which is deployed on the grid platform then. The attention focus of the research is on how to shield the differences among heterogeneous grid services and how to provide a flexible way to compose services. The existing mechanism of grid workflow management usually uses static form, which achieves the aim to compose heterogeneous grid services. But it lacks of flexibility, stability as well as necessary management of atomic services.Heterogeneous grid workflow management mechanism based on virtual service complies with BPEL4WS standards and shields the differences among heterogeneous grid services. The mechanism compensates for the weakness that traditional mechanism requires prior static binding in the service selection and scheduling. Service backup, dynamic service selection based on QoS and so on are all new functions that improve the flexibility of workflow management. In addition, in order to address to the issues of tight coupling between data and service, the mechanism employs virtual data space and relevant data transferring solution. System VSFlow, designed based on such mechanism of service virtualization, integrates the service resources of CGSP (ChinaGrid Support Platform) and GOS (Grid Operating System) and adds functions such as error detection and dynamic service scheduling to the management of virtual service. By the way, the system reserves interfaces for new grid systems to join in, which improves the scalability of VSFlow.The results show that the VSFlow can compose and select grid services on CGSP and GOS, dynamically as well as transparently. By the way, the response time overhead brings by the service virtualization and related management functions, controlled basically between 146.5ms and 156.0ms, has nothing to do with the complexity and time cost of grid service, which does not make system performance decline markedly.