Research on Performance Optimization for Grid Application Using Distributed File System

Applications can be submitted as jobs in grid. Grid jobs usually comprise the scenarios of file stage-in, execution and file stage-out. These scenarios always run in batch mode. It leads to the waiting of the computational resource for data transfer and limits the performance of the grid application, especially when the application is data/communication intensive. Utilizing the file sharing function of distributed file system, the scenarios of grid job can be processed in parallel, while the waiting of computational resource can be eliminated. Thus, the execution performance of the grid job is promoted. After the comparison of different kinds of distributed file systems, this paper proposes CIFS-based mode for the data exchange in grid job, and testifies the promotion of the performance under this mode. In the test of a series of data/communication intensive jobs, the CIFS-based mode performs better performance than GridFTP.