GraphLib: A Parallel Graph Mining Library for Joint Cloud Computing

Graph algorithms are widely applied in social networks, computational biology, Internet security and a broad range of complexity science. Although there are many state-of-the-art graph frameworks, few frameworks support parallel graph mining in joint cloud computing environment. In this paper, we propose GraphLib, a parallel graph mining library, based on a BSP (Bulk Synchronous Parallel) service over joint cloud computing which was proposed in our prior work. We first summarize the features of commonly-used graph mining algorithms, and present our approaches for parallelizing typical graph mining algorithms. GraphLib includes 17 parallel graph mining algorithms that can be used in 3 scenarios. We evaluate the performance of 4 typical parallel graph algorithms in GraphLib on three real-world datasets. Our parallelized algorithms can achieve sub-linear scalability.