Abstract

There are many large-scale graphs in real world such as Web graphs and social graphs. The interest in large-scale graph analysis is growing in recent years. Breadth-First Search (BFS) is one of the most fundamental graph algorithms used as a component of many graph algorithms. Our new method for distributed parallel BFS can compute BFS for one trillion vertices graph within half a second, using large supercomputers such as the K-Computer. By the use of our proposed algorithm, the K-Computer was ranked 1st in Graph500 using all the 82,944 nodes available on June and November 2015 and June 2016 38,621.4 GTEPS. Based on the hybrid BFS algorithm by Beamer (Proceedings of the 2013 IEEE 27th International Symposium on Parallel and Distributed Processing Workshops and PhD Forum, IPDPSW ’13, IEEE Computer Society, Washington, 2013), we devise sets of optimizations for scaling to extreme number of nodes, including a new efficient graph data structure and several optimization techniques such as vertex reordering and load balancing. Our performance evaluation on K-Computer shows that our new BFS is 3.19 times faster on 30,720 nodes than the base version using the previously known best techniques.

Details

Title
Efficient Breadth-First Search on Massively Parallel and Distributed-Memory Machines
Author
Ueno Koji 1 ; Suzumura Toyotaro 2 ; Maruyama Naoya 3 ; Fujisawa Katsuki 4 ; Matsuoka Satoshi 5 

 Tokyo Institute of Technology, Tokyo, Japan (GRID:grid.32197.3e) (ISNI:0000 0001 2179 2105) 
 IBM T.J. Watson Research Center, Westchester County, USA (GRID:grid.410484.d) (ISNI:0000 0004 0400 2468) 
 RIKEN, Kobe, Japan (GRID:grid.7597.c) (ISNI:0000000094465255) 
 Kyushu University, Fukuoka, Japan (GRID:grid.177174.3) (ISNI:0000 0001 2242 4849) 
 Tokyo Institute of Technology/AIST, Tokyo, Japan (GRID:grid.177174.3) 
Pages
22-35
Publication year
2017
Publication date
Mar 2017
Publisher
Springer Nature B.V.
e-ISSN
2364-1541
Source type
Scholarly Journal
Language of publication
English
ProQuest document ID
2407555859
Copyright
© The Author(s) 2017. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.