Parallel Computing Technology Based On ProActive

The high performance computing technology is one of important symbols of national competitive power and scientific researching ability, all countries pay more attention to parallel computing technology. In science and engineering calculation, many questions rely on the parallel computing technology. But the parallel computing technologys at present mostly use MPI/PVM, which have shortcomings such as not supporting heterogeneous environments, unable portable, defficultly scaleable, deployment complex and so on[l,2].ProActive is a pure Java middleware for parallel, distributed, and concurrent computing, which was developed by France's INRIA. With ProActive , user can conveniently develop under the grid and the network environment application[3]. Based on the research of the ProActive middleware, The key technology including the architecture of system, nodes dynamic deployment, parallel programming method based on ProActive and scheduler of parallel tasks, was discussed and designed in detail, a prototype system PHPC (ProActive-based High Performance Computing) was realized. Experiment demonstrates it overcome the shortcomings of current parallel computing, supporting heterogeneous environments, portable, scaleable, simple deployment, and effective. This paper provided theorys and technology for reseach on ProActive-based parallel computing platform.Main works of this paper as follows:1. Given a comprehensive introduction to main technology of the ProActive middleware;2. Based on the research of existed high performance computing system and ProActive, proposed a six layer architecture of ProActive-based parallel computing platform, and given the concrete solution of key technology such as: node deployment, sevices release and so on;3. Study the ProActive parallel programming technology, design application interface of this platform and realized three parallel applications: PI, TSP, and Matrix;4. Considering ProActive's character, design and realize scheduler with pool, static and dynamic scheduling algorithm; Experiment demonstrates it is feasible and...