Research Of Extended Data Management Based On Drivers Under Grid Environment

Most of the data managements of existing grid middleware do the development independently and use one or more of their unique storage and transmission mechanisms, which makes it hard to extend the types of data resources. The existences of data resources are in varied forms and they are maintained by different organizations in different ways. These forms are not always the way new grid applications desire. We want to keep the original way of the existence of data for normal use and also make these resources be used by new grid applications in their way. So providing a highly efficient and scalable model for data management becomes necessity.As one of core modules of ChinaGrid Support Platform (CGSP), the data management module provides high-performance distributed data services, maintains the various heterogeneous storage resources in Grid environment. The current data module of CGSP only supports the GridFTP file system, in order to add the support of new type of data resources, you have to modify the specified module sources according to the given data management system. In addition, the inter-operation with other grid data management systems requires the modifying of sources to make users and applications of CGSP access the data and storage resources of other platforms transparently.The extended data management based on drivers simulates the driver model to provide the function of dynamic adding new types of data resource into CGSP data module. The developers only need to write the drivers in accordance with the specified regulations, which make the data module of CGSP support another type of data resource, such as ftp, http and so on. It also solves problems existed in the current inter-operation between different data management. Developers only have to write some simple drivers so that users and applications of CGSP could access the data and storage resources of grid data management such as SRB and GOS. The hot-swap mechanism makes the drivers join and leave very conveniently, the metadata could be imported into the data space without the restart of CGSP container.Through the actual test, the extended data management based on drivers contains the functions goaled and can add other types of data resources into CGSP data module. At the same time, according to the result of simulation, accessing other data management system through CGSP doesn't affect the efficiency obviously.