In a dedication ceremony held earlier today (March 29), officials from Ohio Supercomputer Center (OSC) along with state representatives gathered to celebrate the launch of OSC’s newest cluster: “Owens.” A nod to the system’s speed and power, “Owens” is the namesake of J.C. “Jesse” Owens, who won four gold medals at the 1936 Olympics. Along with recent upgrades, the new system increases the center’s total computing capacity by a factor of four and its storage capacity by three.
The Dell/Intel cluster was funded as part of a $12 million appropriation included in the 2014–15 Ohio biennial capital budget. An investment of $9.7 million went toward the cluster and the remainder of the appropriation funded storage systems and facilities upgrades, required to support existing and new resources.
“The state of Ohio has made significant investments in the OSC since its creation to expand research in academia and industry across the state through the use of high performance computing services,” said Chancellor John Carey, of the Ohio Department of Higher Education. “Deploying this new system at the center will give Ohio researchers a powerful new tool that they can leverage to make amazing discoveries and innovative breakthroughs.”
As shared in an announcement from OSC, other speakers included David Hudak, Ph.D., interim executive director of OSC; Thomas Beck, Ph.D., professor of chemistry at the University of Cincinnati and chair of the center’s Statewide Users Group; and Tony Parkinson, Vice President for NA Enterprise Solutions and Alliances at Dell EMC.
“This major acquisition, installation and deployment will enable our clients, both academic and industrial, to significantly enhance their computational work,” Hudak said. “Ohio researchers are eager for this massive increase in computing power and storage space. Our current systems were almost constantly running near peak capacity.”
“OSC is dedicated to keeping users involved in the evolution of its HPC systems,” Beck said. “This commitment ensures that research projects are computationally on par with work being conducted by colleagues and partner organizations throughout the state, across the country and internationally.”
The Owens cluster is comprised of a total of 824 Dell PowerEdge server nodes. 648 of these are “dense nodes”: C6320 two-socket servers with Intel Xeon E5-2600 v4 processors and 128GB memory. The system spec page also lists an analytics complement, comprised of 16 huge memory nodes (Dell PowerEdge R930 four-socket server with Intel Xeon E5-4830 v3 processors, 1,536 GB memory, and 12 x 2TB drives).
The theoretical peak performance of the CPU nodes is ~750 teraflops, but the recent addition of 160 Nvidia Pascal-based GPU nodes (Dell PowerEdge R730 two-socket servers with Intel Xeon E5-2680 v4 CPUs), doubles this to 1.5 petaflops double-precision performance — nearly 10X greater any previous OSC system.
The system relies on DDN storage componentry, including Infinite Memory Engine, and Mellanox EDR (100Gbps) InfiniBand.
“OSC’s Owens Cluster represents one of the most significant HPC systems Dell has built,” said Dell EMC’s Parkinson.
The announcement from OSC notes “a nearly complete overhaul of the data center infrastructure has been completed since last spring, now providing users with nearly 5.5 petabytes of disk storage and more than five petabytes of tape backup. The center also acquired and installed NetApp software and hardware for home directory storage.”