Data Management Research In Grid Interoperation System

Grid is a new way to share the distributed resources on the Internet. It organic connects the geographically distributed computing resources,storage resource and information resources so that it provides users with a uniform, standard grid services. Now, because there haven't existing standard protocol of grid and each grid system designed to implement its different utilities, the architecture of these grid is huge different and the resources can not be shared for different grid user. The research of grid interoperation is just aims at these difficulties. The final target is that different grid system can connect with each other and share their distributing resource in a uniform way.Data management plays a very important role in the research of Interoperability on Grid Middleware. It combines different grid system of data resources and provides users and other functional components with a transparent distribution of shared security and fault-tolerant data depository. Each grid has its own data manage module to store the data produced or consumed by grid service. There are two common middleware used in the Data Grid to manage the data, SRB (Storage Resource Broker) and SRM (Storage Resource Manager) and most grids attempt to support either SRB or SRM standard for managed storage. The data interoperation research is to share the data in different grid.In the grid interoperation project, we designed a plug-in interface agent and relying on the strategy to achieve the interoperability between CGSP and SRB. The interface agent we developed can visit both metadata resource and physical data resource in two grid system .It shields the difference of data manage domain and provides users a uniform data view and operation functions with interface-mapping mechanism. Also, it makes sure the data transfer available and consistent by fault-tolerant management.Data resource scheduling arithmetic is the core arithmetic in the grid interoperation system. A good data resource scheduling arithmetic can reduce the time of response and improve data transfer efficiency. This project adopts the interface-mapping scheduling strategy which can choose the metadata path in each grid system and mapping the data operate interface automatically. The experiment results show that the interface-mapping Scheduling has the advantages of quick- response,low-delay and easy-extensibility.