In modern graph analytics, the shortest path is a fundamental concept. Numerous \rrev{recent works} concentrate mostly on the distance of these shortest paths. Nevertheless, in the era of betweenness analysis, the counting of the shortest path between $s$ and $t$ is equally crucial. \rrev{It} is \rev{also} an important issue in the area of graph databases. In recent years, several studies have been conducted in an effort to tackle such issues. Nonetheless, the present technique faces a considerable barrier to parallel due to the dependencies in the index construction stage, hence limiting its application possibilities and wasting the potential hardware performance. To address this problem, we provide a parallel shortest path counting method that could avoid these dependencies and obtain approximately linear index time speedup as the number of threads increases. Our empirical evaluations verify the efficiency and effectiveness.
翻译:在现代图形分析学中,最短的路径是一个基本概念。 许多 rrev{ recent work} 大多集中在这些最短路径的距离上。 然而,在介质分析的时代,计算美元和美元之间的最短路径同样至关重要。\rrev{It} 也是在图形数据库领域的一个重要问题。近年来,为了解决这些问题,已经进行了几项研究。然而,由于指数构建阶段的依赖性,目前的方法面临着相当大的平行障碍,因此限制了其应用可能性并浪费了潜在的硬件性能。为了解决这一问题,我们提供了一种平行的、最短路径计数方法,可以避免这些依赖性,并随着线条数的增加而获得大约线性指数的加速时间。我们的经验评估证实了效率和有效性。