STATE UNIVERSITY AT BUFFALO UB CCR's Industry Cluster Center for Computational Research
UB CCR's Industry Cluster What is HPC NY? What is UB CCR? The UB CCR HPC NY team The UB CCR industry cluster Cluster access Example industry partners Where to go for more information
What is HPC NY? HPC NY is New York State s High Performance Computing Consortium. A network of university computing centers who partner with industries throughout the state to help foster business growth and process improvement. An HPC NY partnership can help companies create jobs, save costs, accelerate R&D, and obtain funding. HPC NY provides access to computational resources and world class expertise in modeling, visualization, and analytics. Funded by ESD/NYSTAR.
HPC NY : Computational and intellectual breadth HPC NY expertise and facilities are distributed throughout the state and linked by the New York State Education and Research Network (NYSERNet): University at Buffalo Rensselaer Polytechnic Institute Stony Brook University Marist College Mount Sinai Powered by ESD/NYSTAR
What is UB CCR? UB CCR is the University at Buffalo's Center for Computational Research. A leading academic supercomputing facility that maintains an HPC machine room and a high-end visualization laboratory. Facilities include a Linux cluster with more than 11,000 processor cores. Remote visualization services provide high-performance graphics tunneling for GUIdriven applications. As part of HPC NY, UB CCR provides a dedicated 72 teraflop ($1.2 Million) cluster for use by industry clients.
The UB CCR HPC NY team A team of UB faculty and staff with expertise in computational science, engineering, scientific programming, data analysis and database design, animation and visualization, and marketing. Research interests include: genomics, bioinformatics, water resources management, molecular modeling, computational chemistry, crystallography, volcanology, disaster response and recovery, computer vision, and computational fluid dynamics. genomics bioinformatics molecular modeling fluid dynamics
The UB CCR industry cluster 216 nodes (Intel Xeon E5-2650v2) 2.60 GHz 16 processors per node 64 GB RAM per node 333 GFLOPS per node peak 72 TFLOPS total 144 with Mellanox FDR IB (56 Gbps) CentOS 7.0.1406 SLURM 14.11.6 Simple Linux Utility for Resource Management Dedicated SLURM controller and database Industry cluster separate from academic cluster (ub-hpc) Industry Cluster Node Rack
Cluster Access Two modes of access: Linux command line via xterm and ssh Remote desktop via web browser client General Procedure Connect to "front-end" Request compute nodes Load software Launch GUI or batch job user s PC UB on-campus or UB VPN compute nodes compute nodes compute nodes presto front-end machine (or remote desktop node) compute nodes
Example Industry Partners Sentient Science Predicting mean time between replacement (MTBR) of wind turbines Embarrassingly parallel Monte Carlo simulations to characterize MTBR under various conditions
Example Industry Partners Buffalo BioBlower A BioBlower is a compressive heating system that kills micro-organisms. It also eliminates chemical toxins and environmental pollutants. The design of the next-generation product is being aided by simulations run on the industry cluster. Cut Away of a Trailer Containing a 5,000 CFM BioBlower Device
Example Industry Partners IBC Digital Nationally recognized animation and visual effects studio Requires a large amount of computing cycles to render and produce final animations. Past Clients: MTV2 Grammy Awards NHL Charlie the Tuna
Where to go for more information Explore the HPC NY Web Site www.hpc-ny.org Contact an HPC NY outreach coordinator: Shawn Matott, UB CCR Computational Scientist and Environmental Engineer lsmatott@buffalo.edu 716-881-7566 Explore the UB CCR Web Site www.buffalo.edu/ccr Contact UB CCR Adrian Levesque, UB CCR Urban Visualization and Multimedia Specialist apl3@buffalo.edu 716-881-8932 Tom Furlani, Director, furlani@buffalo.edu, 716-881-8939 Matt Jones, Associate Director, jonesm@buffalo.edu, 716-881-8958 CCR Helpdesk, ccr-help@buffalo.edu, www.buffalo.edu/ccr/support/ccr-help.html