부산대학교 도서관

In graph analysis area, the shortest path is vital, and recent research shows that the shortest paths counting is crucial in applications like potential friend recommendation and betweenness analysis. Nevertheless, the existing works mainly focus on how to speed up it in a single machine with single core, which do not consider the scalability of it. It limits applications and wastes potential performance. The main bottleneck is dependency between their index is no considered. To fill this research gap, we provide a parallel method. The main approach is to release the dependency between the index as well as optimizations during the process. Moreover, our method could achieve a nearly linear speedup with the number of threads increase in terms of index time. The experimental results demonstrate the effectiveness and efficiency of our method than the baselines.