GPOP: a cache and memory-efficient framework for graph processing over partitions

Abstract

Graph analytics frameworks, typically based on Vertex-centric or Edge-centric paradigms suffer from poor cache utilization, irregular memory accesses, heavy use of synchronization primitives or theoretical inefficiency, that deteriorate overall performance and scalability. In this paper, we generalize the partition-centric PageRank computation approach [1] to develop a novel Graph Processing Over Partitions (GPOP) framework that enables cache-efficient, work-efficient and scalable implementations of several graph algorithms. For large graphs, we observe that GPOP is upto 19× and 6.1× faster than Ligra and GraphMat, respectively.