CoCoS: Fast and Accurate Distributed Triangle Counting in Graph Streams

Given a graph stream, how can we estimate the number of triangles in it using multiple machines with limited storage? Specifically, should edges be processed and sampled across for rapid accurate estimation? The count (i.e., cliques size three) has proven useful numerous applications, including anomaly detection, community link recommendation. For triangle counting large dynamic graphs, recent work focused largely on streaming algorithms distributed but little their combinations “the best both worlds.” In this work, propose CoCoS , fast algorithm estimating counts global all triangles) local incident to each node. Making one pass over input carefully processes stores so that redundant use computational storage resources is minimized. Compared baselines, is: (a) accurate: giving up smaller estimation error; (b) : {10.4\times faster, scaling linearly stream; (c) theoretically sound yielding unbiased estimates.