SC Conference - Activity Details

SGRACE: Parallel Data Management System for Massive-scale Streaming Graphs

Andy Yoo  (Lawrence Livermore National Laboratory)
Posters Session
Tuesday,  05:15PM - 07:00PM
Room Rotunda Lobby
Analyzing terabyte-scale graphs has become an increasingly prevalent and important problem. The irregular nature of the graph analysis algorithms combined with the sheer volume of data sets requires highly scalable data management system that supports the storage and retrieval of massive streaming graphs and efficient execution of out-of-core graph algorithms. We are developing a scalable parallel streaming graph data management system, called Streaming Graph Clustering Engine (SGRACE) to answer these challenges. This report describes the SGRACE system architecture and presents its performance measured using using an out-of-core graph search benchmark on different storage devices. The results indicate that the SGRACE architecture enables the implementation of efficient out-of-core graph algorithms by reducing the number of I/Os and latency.
   IEEE Computer Society  /  ACM     2 0   Y E A R S   -   U N L E A S H I N G   T H E   P O W E R   O F   H P C