Hardware Infrastructure for Network Analytics on Very Large Graphs
Abstract
Network science is becoming a critical instrument for studying complex systems in a broad set of disciplines, such as social network analysis, statistical mechanics, control theory, molecular biology, and neuroscience. However, the field has only barely scratched the surface and really big networks of billions of nodes remain unanalyzed. Key limitation is the lack of network analytics systems that utilize modern computing architectures to its full potential. We propose to acquire a state of the art big-memory machine with multiple terabytes of memory and hundreds of cores. The machine will allow us to develop next-generation network analysis systems that to full-extent exploit such modern computing architectures and perform network analytics significantly faster than traditionally used distributed systems. This experimental machine will give us low level system access, required to do research at a fundamental level and not available in production environments. Our objective is to expand the scope of network analytics to very large real-world graphs with billions of nodes and tens of billions of edges or more. These graphs occur commonly in a range of fields. Examples are social network analysis with over 7 billion of living people, neuroscience with more than 100 billion neurons in human brain, biology with 3 billion DNA pairs in a single human or statistical mechanics with over 1020 atoms in a cubic mm of diamond. DURIP is a unique venue to fund proposed infrastructure, since its cost is beyond the reach of typical grants that we are normally awarded by DoD agencies. The infrastructure will be utilized by ongoing DoD funded research programs in Prof. LeskovecÕs group at Stanford University. As the group attracts many students at PostDoc, PhD, MS and BS levels, the infrastructure will facilitate education of young researchers in advanced network science methods.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 11, 2018
- Source ID
- W911NF1610171
Entities
People
- Jure Leskovec
Organizations
- Army Contracting Command
- Stanford University
- United States Army