Implementation of a parallel graph partition algorithm to speed up BSP computing

The processing and mining information in large scale graph data have proven to be challenging. The bulk synchronous parallel (BSP) computing model is suitable for this task. In this paper, we implement the multi-level step-wise partitioning (MSP) algorithm in BSP programming model, and replace the original graph partition method. The results on both experimental data and real world data proved this improvement achieved better data locality, reduced communication between work nodes, and it made a better performance than the original method.