In the battle against the COVID-19 coronavirus, we need all our weapons to defeat this enemy. And high-performance computing (HPC) may be one of the most valuable weapons. An important part of treating this virus will be understanding it biological interaction and modeling various treatments. This is where the power of HPC to model chemical reactions can help.
The U.S. Federal government has quickly assembled a group of HPC related resources and put them at the disposal of researchers looking into treatments and cures for the COVID-19 virus (the White House statement). The government program is called the COVID-19 High Performance Computing Consortium and it brought together the U.S. government, industry, and academia to provide access to the powerful high-performance computing resources. The consortium has access to over 366 petaflops on total compute power with 2,839,772 CPU cores and 36,058 Nvidia GPUs and more resources are being added.
The initial industry leaders participating in the program include: Amazon Web Services, Google Cloud, Hewlett Packard Enterprise, IBM, and Microsoft. Even though the program just started, AMD and NVIDIA have quickly jumped on board as well. In addition, the U.S. Department of Energy (DOE) has several National Laboratories that are contributing resources and compute power including: Argonne National Laboratory, Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Oak Ridge National Laboratory, and Sandia National Laboratories. The other Federal agencies involved include: the National Science Foundation (NSF) and NASA. The initial academic institutions participating included both Massachusetts Institute of Technology and Rensselaer Polytechnic Institute. Many more universities and labs have joined since the initial announcement (see a complete list at the consortium website).
The consortium hopes that by providing access to this massive amount of computing resources, the researcher scientists can process numbers of calculations related to bioinformatics, epidemiology, and molecular modeling to derive answers to complex scientific questions about COVID-19 “in hours or days versus weeks or months.”
Oak Ridge National Lab Summit Supercomputer
The DOE has granted researchers emergency access to the Summit supercomputer at Oak Ridge National Labs (ORNL). Running simulations for two days on Summit identified and studied 77 small-molecule drug potential compounds to fight against the COVID-19. Using a traditional “wet-lab” approach to the research could take years. The ONLR researchers performed simulations of more than 8,000 possible compounds to screen for those that bind to the main “spike” protein of the coronavirus, rendering it unable to infect new host cells. Viruses use a ‘spike’ to inject their genetic material into a host cell. While the final test is done in a wet lab, digital simulations can narrow down the range of potential compounds to test. Summit is a 200 PetaFlop supercomputer composed of more than 9,000 IBM Power9 CPUs, connected to more than 27,000 NVIDIA Volta GPUs.
NVIDIA’s Response To COVID-19
As a leader in high-performance computing (it’s the GPU provider for Summit), NVIDIA has a number of tools it can offer to help researchers. As mentioned earlier, the company has joined the COVID-19 High Performance Computing Consortium and will provide its expertise in AI, accelerated computing for science, and computer system optimization expertise. Ian Buck, Nvidia’s vice president and general manager of Accelerated Computing is leading the team.
Nvidia was already deep into genome analysis and it bought gene sequencing software company Parabricks in December. Parabricks is another tool that can accelerate research, getting results in days rather than weeks. The Parabricks genome analysis toolkit uses GPUs to accelerate analysis of gene sequencing data by up to 50x. The help researcher get access to the tool, NVIDIA will provide a free 90-day license to any researcher working to fight the novel coronavirus.
NVIDIA is also bringing on partners that can offer additional cloud services and supercomputer centers to run Parabricks. Those partners include Oracle, Core Scientific and Net App. Oracle is providing NVIDIA GPU Cloud (NGC) machine images through Oracle Cloud Infrastructure, Core Scientific is providing free access to NVIDIA DGX systems, and NetApp is providing cloud-connected storage through ONTAP AI. Tencent Cloud will also provide free GPU instances up to 90 days for qualified scientific research institutions.
PC Gamers Join The Fight
But you don’t have to have access to a supercomputer in order to help the fight. You can also help if you have a high-performance computer at home for gaming or content creation. The Folding@Home project has been around for many years and it breaks problems into smaller workloads that can be run on a distributed network of individual PCs. The Folding@Home program performs molecular dynamics simulations of protein dynamics by breaking up the compute load into smaller tasks and sending those tasks out to individual PCs running the client software. Each volunteer PC completes its part of a larger simulation and then sends it back to the project’s database server where they are combined into the overall simulation.
The combined performance of the many Folding@home clients represents a total computations capability that rivals the world's fastest supercomputers. Interest in the project has jumped as a result of the novel coronavirus pandemic, and the total system achieved a speed of approximately 768 petaFLOPS, or 1.5 x86 exaFLOPS, on March 25, 2020. By March 30th, the system included over 356,000 Nvidia GPUs, over 79,000 AMD GPUs, and over 593,000 x86 CPUs.
Conclusion
The GPU has proved to be a powerful weapon in accelerating many workloads. While the GPU are a great tool to fighting imaginary enemies in computer games, it now has a key role in the real-life fight against the COVID-19 virus. NVIDIA, with its deep investment in HPC software and systems is extremely well equipped to support this effort.
Final note: Because of the pandemic, NVIDIA has also been forced to turn its yearly GPU Technology Conference (GTC) into a streaming event. But it has also extended the program over multiple weeks and put much of the content online after the sessions are over. Check out more of GTC here.
Kevin Krewell
Sheltering in place.
