The STAPL Parallel Graph Library
November 1, 2012
2:30 p.m.
Harshvardhan
Abstract
Processing large graphs is essential to many domains, from social network and web-scale graphs to scientific meshes and nuclear reactor-design. As the graphs span billions of vertices and edges, they may not fit in the memory of a single-processor system. Using a distributed data-structure allows massive graphs to be processed quickly and concurrently. However, graph algorithms remain notoriously hard to parallelize, and existing solutions to address this need expose too many details about parallelism, data-distribution and communication.
In this work, we present the STAPL Parallel Graph Library, an expressive and customizable high-level framework that abstracts the user from data-distribution and parallelism details and allows them to concentrate on parallel graph algorithm development. Experimental results demonstrate improved scalability in performance and data size over existing graph libraries on more than 16,000 cores and on internet-scale graphs containing over 16 billion vertices and 250 billion edges.