Large-scale graph analytics in Aster 6: bringing context to big data discovery

• Published in: Proceedings of the VLDB Endowment 2014-08, Vol.7 (13), p.1405-1416
• Main Authors: Simmen, David, Schnaitter, Karl, Davis, Jeff, He, Yingjie, Lohariwala, Sangeet, Mysore, Ajay, Shenoi, Vinayak, Tan, Mingfeng, Xiao, Yu
• Format: Article
• Language: English
• ISSN: 2150-8097; 2150-8097
• DOI: 10.14778/2733004.2733013

Graph analytics is an important big data discovery technique. Applications include identifying influential employees for retention, detecting fraud in a complex interaction network, and determining product affinities by exploiting community buying patterns. Specialized platforms have emerged to satisfy the unique processing requirements of large-scale graph analytics; however, these platforms do not enable graph analytics to be combined with other analytics techniques, nor do they work well with the vast ecosystem of SQL-based business applications. Teradata Aster 6.0 adds support for large-scale graph analytics to its repertoire of analytics capabilities. The solution extends the multi-engine processing architecture with support for bulk synchronous parallel execution, and a specialized graph engine that enables iterative analysis of graph structures. Graph analytics functions written to the vertex-oriented API exposed by the graph engine can be invoked from the context of an SQL query and composed with existing SQL-MR functions, thereby enabling data scientists and business applications to express computations that combine large-scale graph analytics with techniques better suited to a different style of processing. The solution includes a suite of pre-built graph analytic functions adapted for parallel execution.