GPOP: A cache- and work-efficient framework for Graph Processing Over Partitions

Past decade has seen the development of many shared-memory graph processing frameworks intended to reduce the effort of developing high performance parallel applications. However, many of these frameworks, based on Vertex-centric or Edge-centric paradigms suffer from several issues such as 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 a recent partition-centric paradigm for PageRank computation to a novel Graph Processing Over Partitions (GPOP) framework that exploits the locality of partitioning to dramatically improve the cache performance of a variety of graph algorithms. It achieves high scalability by enabling completely lock and atomic free computation. Its built-in analytical performance model enables it to use a hybrid of source and partition centric communication modes in a way that ensures work-efficiency each iteration while simultaneously boosting high bandwidth sequential memory accesses. Finally, the GPOP framework is designed with programmability in mind. It completely abstracts away underlying programming model details from the user and provides an easy to program set of APIs with the ability to selectively continue the active vertex set across iterations. We extensively evaluate the performance of GPOP for a variety of graph algorithms, using several large datasets. We observe that GPOP incurs upto 8.6x and 5.2x less L2 cache misses compared to Ligra and GraphMat, respectively. In terms of execution time, GPOP is upto 19x and 6.1x faster than Ligra and GraphMat, respectively.