And boosts to other U.S. supercomputing systems are likely on the horizon.
A Lawrence Livermore National Laboratory-based supercomputer driving coronavirus-centered research received its second upgrade since the beginning of the pandemic.
Named not for the virus but after the 2017 total solar eclipse, the National Nuclear Security Administration’s Corona high performance computing system recently received assets from chipmaker AMD and information technology company Supermicro that more than double its speed and capacity.
Such a boost will pave the way for more robust compute cycles and improve how the supercomputer handles deeply complex simulations—and it was made possible in part by funding from the government’s initial coronavirus relief package, the CARES Act.
“While we aren’t releasing the cost of the latest upgrade, we can say that LLNL received a total of $8.7 million in CARES Act funds to buy computing equipment for COVID-19 related research, including this latest expansion to Corona and several other systems that will be announced in the next few weeks,” the lab’s Public Information Officer Jeremy Thomas told Nextgov Thursday. “The funds came through the [Energy Department] Office of Science’s Advanced Scientific Computing Research program.”
Between its delivery in 2018 and the emergence of COVID-19, Corona was predominantly used to drive forward unclassified, open science applications. And like other key U.S. supercomputing systems, when the virus hit, Corona was quickly pivoted to power coronavirus-related work, particularly in the realm of drug discovery and vaccine development. Early into the pandemic, the system went through an initial upgrade that included accelerators from AMD optimized for large-scale deep learning. The boost enabled Corona to reach 4.5 petaFLOPS—or floating point operations per second—of peak compute power.
The latest upgrade unveiled Wednesday equips the system with another nearly 1,000 AMD accelerators, as well as motherboards, servers and racks necessary for the expansion that were provided by Supermicro. Those resources “more than double the speed of the cluster from about 4.5 petaFLOPS to in excess of 11 petaFLOPS,” according to the lab’s announcement on the work. In it, LLNL Deputy Associate Director for Programs Jim Brase said “this performance boost will help the Corona system lead the way in accelerating pandemic response,” and Brad McCredie, corporate vice president, Data Center GPU and Accelerated Processing at AMD added “we aim to help researchers find creative ways to fight or cure COVID-19 much sooner.”
That fully-expanded Corona system is now already in use and has been for several weeks.
“The installation of the new hardware took about two weeks to complete, with the initial Corona system remaining in use during that time,” Thomas told Nextgov. He further added, “the expanded Corona system is being used for a massive number of ensemble simulations, with the number of total jobs run already in the thousands and increasing rapidly.”
Thomas further detailed some of those pursuits, including discovery and design of antibodies that might bind to SARS-CoV-2 (the virus that causes COVID-19). The system is also being tapped to virtually screen more than a billion small molecule compounds to help scientists discover therapeutics and antiviral drugs. And on top of those LLNL-led efforts, researchers worldwide can also tap into the system through the COVID-19 HPC Consortium, launched previously by the White House Office of Science and Technology Policy, Energy Department and IBM.
In the release, Brase, who also heads Lawrence Livermore’s COVID-19 research and rapid response effort, called the Corona system “a major advance in our capability for predictive biomedical modeling for COVID-19.”