Design And Implementation Of Cluster Computing System Based On MPI

The technology of computer has prompted the rapid growth of the computational science. The new generation of computer should be better than the former either in the power or the speed of computing. Furthermore, the demand of computing is increasing. However, in practice single processor still can't satisfy the demand of some engineering calculation. In addition to improving the computing capability of processor, parallel computing is also a potent way to increase the capability of computing. The parallel computing can be well implemented in cluster computing system, which is necessary to organize several computers to work together in order to get powerful parallel computing. These heavy tasks are distributed to all of existing computation resources by cluster computing system.In the thesis, the structure and function of Linux cluster are firstly discussed in details, and then the state of cluster and the parallel programming model in parallel computing are analyzed. It is researched on the single system image and the cluster middleware. A new type of parallel cluster system is designed using cluster middleware. After implementing functions of some modules in cluster system, this system has a testing capability for network overload, and it can dynamically reflect the change features for nodes and rationally distribute tasks, according to the load of node. These researches and implementations are pop research subjects on parallel computing system, and provide references for relative researches.At the end of this thesis, two parallel applications, using static dispatching algorithm and dynamic weighted round-robin dispatching algorithm separately, are designed based on this cluster system. The external performance of the cluster system is tested using client application, and then testing result is analyzed. The experiment indicates that the cluster system has excellent computing performance and rapid response to the external performance.