Study Of Transplanting MPI Applications In Grid Environments Through The Example Of The Mesoscale Numerical Weather Prediction Model

Grid computing is developed with the internet technologies. It is used to make use of computing resources which are in different positions such as databases; expensive instuments and so on to solve complicated large-scale problems, especially those that we can not solve depend on the local resources.Grid computing brings us a new high performance computing solution. But at present, grid computing field is in developing and immature. It becomes an important task that how to transplant the existing applications into the grid environments and enjoy the grid services that grid computing and globus toolkit provide.In recent years, the high performance parallel computing applications based on MPI messaging passing model have been widely used. With the incessant scale-up of the problem scope and the incessant increase of the user demand, we need seek help from grid and make use of the services that grid provides such as GRAM and MDS.Based on the characteristics of grid computing and MPI, It has been testified by experiments that it is a practical solution to transplant MPI applications in grid environments. We can not only solve the problem of the programming language in grid environments but conceive the grid environments in the smallest cost.In the experiment we have built a grid computing platform with four nodes, three nodes in linux operating system and one node in solaris operating system. We have done a completely test on the function and capability of the gird. Then we transplanted the the mesoscale numerical weather prediction model MM5into the grid environment. By analyzing the experiment data and evaluating the capability, we get the conclusions that the computing system can be easily extended, has high usability and low spending. It brings us a better basis to research the grid computing technology and build the school gird computing center in the future.