Research of data partitioning technique for parallel load balancing
面向并行负载平衡的数据剖分技术*

This paper reviewed the traditional data partitioning technique and the impact of the load balancing on the large-scale parallel computing performance,described the features of the geometric methods and graph partitioning methods.Compared the advantages and disadvantages of the various partitioning methods and partitioning packages(like ParMETIS,Zoltan,JOSTLE and so on),in order to solve the load balancing problems.Particularly discussed how to efficiently use the data partitioning technique to partitioning the large scale numerical simulation tasks.It is beneficial for those who involve in research of load balancing and parallel optimization.