Study And Design Of Tianjin University's Grid System

With the development of school research and network-based applications,school demands for much more resources.But the distributed and heterogeneous resources of institutes doesn't have the management which carries out effectively and unifies, making the utilization of resources greatly reduced.The traditional distributed computing systems tend to integrate the isomorphic resources, thus limiting the scope of integration of resources and user's applications;while the super-computer as the resource center providing resources is expensive and lacking of scalability,and provides relatively limited application types,thus can not meet user's diversity demands. The grid technology has changed the way people make use of network resources.It can integrate the distributed and heterogeneous resources to make them work together.This paper focuses on grid architecture and modules in the architecture.By analyzing the characteristics of the resources in the campus network,the author proposes a architecture of the campus grid.The paper focuses on the Globus Toolkit (GT),a grid middleware,and proposes the program of implementation of Tianjin University's Grid System,which is scalable,transparent to the users and security to integrate heterogeneous resources to improve the utilization of resources. GT has some deficiencies in job management and scheduling which are not friendly for users. The paper references TeraGrid project to use GX-Map kits as a grid-mapfile management system; proposes a program combining Condor's job management and GT to achieve a friendly operational management; proposes the detail design of information services and take matchmaker component for example to show the problems scheduling component need to address and give some solutions.Campus Grid system mainly serves school's teaching and research,high performance computing.is its main application. Since parallel computing is the basis for high-performance computing,this paper tries to use Globus Toolkit and MPICH to built a grid environment of parallel computing. It has achieved a grid parallel computing environment and resource monitoring by testing.