Parallel I/O Simulator Based On Co-design In The Cluster Computer System

Cluster computer system has notable features: increased scalability, high degreeof accessibility and better performance/price ratio and other. It is widely used in thecomplex environment of the huge data to be treated or calculated in recent years.Cluster system is a high performance parallel computing system. It can solve theproblem that the speed of parallel computing is much different from the access speedof parallel I/O, through using a variety of parallel technology and design techniques.The goal of this project is that the realization of parallel I/O simulator based onco-design in the cluster computer system. On the basis of the original design ofparallel I/O simulator, it can choose the optimal parameters of various parameters inthe simulator to improve the performance of the original simulator. Under theguidance of collaborative design theory, there is no knowledge differences betweendesigners. And also, the function of the original system can be enhanced.The research and analysis of collaborative design theory is done in the paper. Theideas that the realization of parallel I/O simulator based on co-design in the clustercomputer system is proposed based on the theory research. It uses the specificalgorithm of collaborative design and the reformed genetic algorithm to optimize thedesign. At last, it designs a framework which supports parallel I/O simulator in thecluster system. And also it realizes each core modules in the simulator framework. Ituses hash table based on Trie-tree for system-testing. The results indicate that its I/Othroughput and the capacity of parallel computing are greater than before.The reformed parallel I/O simulator based on co-design in the cluster system cansolve the problem of automatic optimization to meet the bottleneck of the clustersystem.