After a lengthy competitive bid process and numerous supply chain delays, UMD announced its latest investment in high-performance computing. Given the name Zaratan, after the mythical sea turtle known for its long lifetime and grandiose size, this new flagship HPC cluster will support world-class research at UMD. The university is dedicated to pushing the boundaries of scientific understanding as it commits to empowering its research community with an accessible, superior leading-edge HPC system.
“We take the enablement of science using cutting-edge resources very seriously, and pushing the envelope of science and discovery is aided by computational resources and other prevailing technologies,” says Tripti Sinha, Assistant Vice President and Chief Technology Officer of the UMD Division of Information Technology. “That’s our role—helping our researchers push the frontiers of science.”
The needs for more compute cores, more data storage and more GPU accelerators formed the foundation for the development of Zaratan. Deepthought2, Zaratan’s predecessor, could provide a theoretical peak performance of 300 teraflops. Zaratan’s 360 liquid-cooled compute nodes, 50,000 CPU cores and 80 GPUs provide more than a ten-fold performance increase over Deepthought2. In addition, Zaratan provides a superior storage capacity for big data sets, with ten times as much storage, including a much-needed archive tier.
“Zaratan embodies UMD’s strategic goals of tackling grand challenges. Solving many of the world’s grand challenges requires advanced computing technology,” said Jeff Hollingsworth, Professor, Vice President and Chief Information Officer at UMD.
UMD aspires to take HPC to disciplines beyond traditional supercomputing domains, and put the power of HPC into the hands of more faculty members, thereby expanding beyond the hard sciences. To help meet that goal, Zaratan offers an Open OnDemand web portal. “Open OnDemand removes many of the complexities that come with accessing HPC systems by traditional methods,” notes Kevin Hildebrand, HPC Architect at UMD. “With Open OnDemand, HPC users have an easy path to get the compute resources they need when they need them, thus lowering the barriers to entry.”
To read the full press release, click here.