Data Management And Transfer Service Research In High Performance Computing

With the development of science and technology, the requirements of the high performance computing are increasingly high. Grid can enable geographically wide area distribution of resources including computing resources and data resources organized in a uniform framework for sharing. How efficient to transfer the massive data is a very important problem in grid computing.Not only security data services must be provided by grid data management environment, but the distributed storage management, data assemble, transfer services, replication management with hotspot data backup and security of data, high-speed data transfer services are also needed. For some massive data, we hope to be able to conveniently and speedily to transmit them to the specified with minimize transfer time. At the same time, we also need to backup hotspot data in multi-processors, namely management the data backup.Through analyzing and studying the grid structure, this paper focuses on data management and transfer services in high performance computing environment. Based on the Globus Toolkit and GridFTP which is file transfer protocol supports grid security architecture, integrated graphics platform of grid data transfer (GridDT) services is designed and realized. GridDT enables files upload and download, parallel data transfer and transfer controlled by the third party. The transfer performance in the graphic and command manner is compared and the stripe data transfer mode is analyzed. At the same time, principles of the replica management are studied, the integrated graphics platform of data management (GridDMC) is realized, which provides replication creating, query of the grid data and the replica location to meet the requirements of the applications.