Distributed Graph Database for Large-Scale Social Computing

Abstract

We present an efficient distributed graph database architecture for large scale social computing. The architecture consists of a distributed graph data processing system and a distributed graph data storage system. We leverage the advantages of both systems to achieve efficient social computing. We conduct extensive experiments to demonstrate the performance of our system. We employ four real-world, large scale social networks - YouTube, Flicker, LiveJournal and Orkut as test data. We also implement several representative social applications and graph algorithms to examine the performance of our system. We employ two main optimization techniques in our system ¡Vindexing and graph partitioning. Experimental results indicate that our system outperforms GoldenOrb, an implementation Pregel model from Google.