The Design And Implementation Of A Metadata Information Service System In Data Grid

In recent years, volume of datasets in modern large-scale scientific researches, information services and digital media applications is growing explosively, and research about data grid technology becomes a new hotspot in the computer circle all over the world. Since 1990 some great progresses have been made about the basic theoretical research and test-bed environment construction in USA and European countries. Researches about the Grid computing technologies have been made.Our project focuses on the key technologies of data grid. According to the characteristic of data intensive application, we design and develop our own data grid called GridDaen. It adopts Java mechanism, provides uniform access and management of the large scale distributed scientific datasets and converts different bottom accessing protocols automatically, providing a service platform for high level application development.This paper studies the meta-information service in data grid, as an important part of the GridDaen project. On the basic of advance grid systems and their meta-information technologies, we design and realize a distributed meta-information system which can support different managing domains, called MDIS. It is completed by using RMI mechanism and JDBC database tools, providing friendly interface and easily used publish & discovery tools for users and APIs & SDK for programming. MDIS provides some important services such as information publish & discovery, replica & management, multiple-domain security mechanism and system management. Therefore, grid users in different locations are provided with several useful features such as uniform logical view, single accessing interface and virtual collection mechanism. It shields the difference between diverse underlying resources and accesses data transparently, which can be a favorable supporting platform to exploit higher grid application.On the basis of the meta information service research and realization above, a self-determinate replica selection policy in Grid environment is carefully analyzed and designed. Testing results indicate that it can really enhance the total performance of the whole system by building up the locality and robustness of data resources.