The Research And Implement Of Computational Grid Prototype GridAG

With the rapid development and wide application of Internet technologies, a new computing pattern, Grid computing, which is booming. Grid is an Internet-based infrastructure for resource sharing and cooperative compute, which brings the Internet into a new era. Grid is a compositive computmg and resource environment, which can contain all kinds of computing resources, and change them to a environment which has reliable, standard, economic computing capability. Computational grid is a important aspect of the research of grid, which can aggregate all kinds of distributed and heterogeneous resources on the network, and change them into relatively transparent, virtual high performance environment.We research and implement the Computational Grid prototype GridAG based on agent, and its goal is couple up the dispersed computing resources into a application environment which has high computing capability and provide computing service for user.The paper analyzes the architecture of Grid firstly, introduces Grid computing technology, Globus and its resource management in detail. Then, introduces CoG and MPICH-G2, and analyzes their application in the GridAG. Based on these analysis and research, this paper presents design and implementation of GridAG Computational Grid prototype, proposes GridAG's architecture and resource information model, and introduces service sharing and security access of GridAG. And this paper also proposes MCT task scheduling algorithm based on grid resource reliability which not only enhances the executive reliability of task, but also achieves load balance among grid nodes. The system can be divided into grid client-side (Web Portal). Resource request agent (the agent-based grid middleware) and grid server-side logically. Grid client-side solves the problem that user use GridAG's resources through uniform interface, Grid server-side executes user's task, Resource request agent is a middleware which connects Grid client-side and Grid server-side. The implementation of the GridAG is service-oriented and modular, so it is flexible, manageable, scalable, high available and so on.