2015 IEEE International Conference on Smart City/SocialCom/SustainCom (SmartCity) (2015)

Chengdu, China

Dec. 19, 2015 to Dec. 21, 2015

ISBN: 978-1-5090-1892-5

pp: 715-720

ABSTRACT

Graph computing is widely utilized today, which severely requires the ability of processing graphs of billion vertices rapidly for social network analyzing, bio-informational network analyzing and semantic processing. Therefore, graph processing play a significant role in the research and application development. Data of music and movie recommendation and LDA topics can be modeled as bipartite graph and perform the computation with graph processing engines. The most important step before graph computation is graph partitioning. Graph partitioning is a mature technology, however, most of classic graph partitioning algorithms require iterative calculation for several times, which causes high time complexity. Some algorithms with short partitioning time proposed these years, but they cannot be used in bipartite graph directly. This paper proposes a new bipartite graph partitioning algorithm, BiFennel, which effectively decreases graph processing time and network loading by reducing vertex replication factor and maintaining work balance. We implement BiFennel in a popular graph engine called PowerGraph. The performance results show that BiFennel has 29~55% improvement on communication cost and 21~49% improvement on overall runtime comparing with Aweto.

INDEX TERMS

Partitioning algorithms, Bipartite graph, Engines, Algorithm design and analysis, Computer science, Computational modeling, Software algorithms

CITATION

L. Wang, S. Chen, W. Chen, H. Hsiao and Y. Chung, "BiFennel: Fast Bipartite Graph Partitioning Algorithm for Big Data,"

*2015 IEEE International Conference on Smart City/SocialCom/SustainCom (SmartCity)(SMARTCITY)*, Chengdu, China, 2015, pp. 715-720.

doi:10.1109/SmartCity.2015.153

CITATIONS

SEARCH