Data Transmission Performance Analysis And Research Based On GridFTP

Data-intensive scientific and engineering applications require both transfers of largeamounts of data between storage systems and access to large amounts of data by manygeographically distributed applications and users for analysis, visualization, etc. So-calledData Grids provide essential infrastructure for such applications. There are a lot of Data Gridresearch projects, but the main open and universal protocol platform is Globus. Globus’shigh-speed transport service, GridFTP, extends the popular FTP protocol and adds newfeatures required for Data Grid applications.Proceeding from the data transmission service in grid, this thesis analyzes themechanisms of GSI security, resources and services management in Globus, and conducts theperformance analysis and research of data transmission on the different transport modes of theGridFTP protocol. First of all, this thesis carries out research with GSI security infrastructureand GridFTP protocol for Globus Toolkit, and uses Globus Toolkit Kit to build a gridenvironment. The GSI and GridFTP protocol are configured on the nodes to meet therequirements of the secure transmission. And this thesis installs and configures the Java CoGKit so that the nodes meet the requirements of the development of grid transmission system.Then, to design the grid transmission system as a whole, by using Java CoG Kit, this thesisimplements the simplified system, including file resources management, securityauthentication module and a variety of transmission modes. Finally, this thesis tested thetransmission respectively in three modes, namely FTP protocol mode, GridFTP protocolcommand line mode and the data transmission system. So the parameters, such as transfertime, throughput and total data size, were got on the different transport modes and parallelismof the GridFTP protocol.The experiments show that the GridFTP protocol has obvious advantages whentransferring large data. Multiple parallel TCP streams reduce the transmission time even whentransferring more data, which increase the transmit throughput, and the transmissionperformance improve more than the original. The grid transmission system developed in thisthesis makes it convenient for users to transfer grid data. The cross-platform data transmissionprovides a reference for the further study of the GridFTP protocol. This research results haspractical significance to the exploring of access to improve performance of mass datatransmission and promoting the application of grid technology.