Comparison of Edge Partitioners for Graph Processing

Deploying graph on a cluster requires its partitioning into a number of subgraphs, and assigning them to different machines. Two partitioning approaches have been proposed: vertex partitioning and edge partitioning. In the edge partitioning approach edges are allocated to partitions. Recent studies show that, for power-law graphs, edge partitioning is more effective than vertex partitioning. In this paper we provide an overview of existing edge partitioning algorithms. However, based only on published work, we cannot draw a clear conclusion about the relative performances of these partitioners. For this reason, we compare all the edge partitioners currently available for GraphX. Our preliminary results suggest that Hybrid-Cut partitioner provides the best performance.