Efficient Parallel Cycle Search in Large Graphs

Cycle is a fundamental structure in graphs. Motivated by the wide applications of cycle search, namely computing the cycles related to a vertex in the graph, we investigate efficient parallel algorithm to address the cycle search problem in large graphs. We first propose a two-phase paradigm tailored for the parallel cycle search problem. Based on the paradigm, we further devise a workload estimation method to improve the efficiency and scalability of the algorithm by balancing the work assigned to different threads. We experimentally evaluate our algorithms on real datasets and the results demonstrate the effectiveness of our approach.