"High-Performance Graph Analytics on Emerging Hardware" presented by Hang Liu, The George Washington University
Abstract:
Graph analytics is of great importance to various disciplines and domains as many real-world datasets come in the format of graphs, to name a few, social networks, knowledge graphs, metabolic interaction networks and road maps. An example that is closely related to everyone’s daily life is route planning (navigation) which employs graph algorithms to compute the estimated shortest path from source to destination and provide guidance for the driver. However, as graphs continue to grow, as well as their irregularity aggravates, mining valuable information in an acceptable time envelope becomes more and more challenging. Generally speaking, graph processing faces four critical challenges -- workload imbalance, random memory access, high computation cost and ever growing huge graph size. In this talk, I will present how my work -- Enterprise (SC '15), iBFS (SIGMOD '16) and Graphene (FAST '17) -- is able to leverage the massive potentials of emerging hardware, i.e., Graphics Processing Unit (GPU) and Solid-State Drive (SSD), to address those challenges and achieve unprecedented performance for graph analytics applications.
Bio:
Hang Liu is a PhD candidate in Department of Electrical and Computer Engineering at The George Washington University. His research lies broadly in Big Data Analytics, Numerical Simulations, Cloud Computing, and High-Performance Computing. His graph work has been ranked highly in Graph500 and GreenGraph 500, e.g., Graphene is the most energy efficient graph traversal in USA (Big Graph Category, June 2016).
Wednesday, January 18, 2017 at 12:20pm to 1:20pm
McAdams Hall, 119
821 McMillan Rd., Clemson, SC 29634, USA
College of Engineering, Computing and Applied Sciences, School of Computing, Research Seminars
No recent activity