Lawrence Livermore National Laboratory, along with its partners AMD and Supermicro, and Cornelis networks have installed a new high-performance computing (HPC), cluster with memory storage and data storage optimized for data-intensive COVID-19 research and pandemic response.

The Coronavirus Aid, Relief, and Economic Security Act (CARES Act) fund the Mammoth “big memory” cluster. It will be used by LLNL to perform genomics analyses, non-traditional HPC simulations, and graph analytics needed by scientists involved in COVID-19, which includes the development of antivirals and designer antibodies.

Jim Brase, LLNL’s deputy assistant director for Computing, stated that large-memory systems will allow them to combine genomic analysis with large-scale machine learning to predict therapeutic responses. This will help accelerate the development of new effective therapeutics. Mammoth will play a vital role in developing new tools to fight COVID-19 and for rapid response in the event of a pandemic.

Mammoth is composed of 64 nodes equipped with second-generation AMD EPYC processors. Each node is equipped with two 64-core CPUs and 128 threads. It also features high-memory bandwidth, 2 terabytes of DRAM memory (TB), and almost 4 TB nonvolatile storage. COVID-19 researchers who need to sift through large databases of information will appreciate the Mammoth’s extra memory.

“It’s exciting to see technology and science directly connected to the improvement or even saving of lives,” Dan McNamara (senior vice president AMD Server) said. AMD is proud to support the important research being conducted by the team at LLNL and Mammoth with the help of our technology partners, Supermicro Networks and Cornelis Networks.

LLNL researchers applied Mammoth to the SARS-CoV-2 virus, which causes COVID-19. They used the cluster to study how the virus evolves. The cluster also simulates how it changes as they introduce mutations. Lab scientists stated that Mammoth has reduced the time required to do certain types of genomic analysis. It takes only a few minutes instead of a few days.

“This new system makes it a big improvement in how we prepare for calculations and their performance,” Adam Zemla, a researcher who uses Mammoth to detect virus mutants, clustering, structural modeling, molecular dynamics simulations, and other related tasks. Mammoth allows me to process COVID-19 genomes quickly without having to break them into smaller pieces, as I did on other machines.

Researchers at LLNL said that Mammoth’s many cores and large memory resources are helping to design modified antibodies that will better bind against the virus.

Thomas Desautels, a computer scientist at LLNL, said that binding-free energies are computed with Rosetta Flex in our workflow. This code was memory-limited on other machines and limited to 12 or 16 simultaneous calculations per machine. Mammoth allows us to run 128 Rosetta Flex calculations simultaneously from a single node. This increases our throughput by about eight. Mammoth allows us to run many more Rosetta Flex calculations simultaneously on a single node, which speeds up our search for SARS-2 antibodies.

Mammoth, an AMD-based production HPC cluster, is connected to components from Cornelis Networks. Cornelis Networks provides purpose-built interconnects that are focused on HPC and high-performance data analytics (HPDA) as well as artificial intelligence (AI). Cornelis Networks, a spinout of Intel Corporation was launched in September with an extensive interconnect portfolio to meet the demands of demanding workloads, as well as future innovative, interconnect products.

Cornelis Networks CEO Phil Murphy stated that “Cornelis Networks was thrilled to partner AMD and Supermicro in delivering Mammoth for the researchers at Lawrence Livermore.” “Accelerating the pathway to a therapeutic for COVID-19 exemplifies our company’s mission to deliver innovative solutions that advance our customers’ missions to solve the most difficult problems in the world,” Murphy said.

Supermicro, a San Jose-based information technology company, provided Mammoth with the hardware, including the motherboards and racks.

Charles Liang (president and CEO of Supermicro) stated that Supermicro is thrilled to support another cluster at LLNL devoted to understanding the genetic makeup of the COVID-19 viruses. The 1U A+ Ultra, which is installed in the Mammoth cluster, leverages the system’s 64-core AMD EPYC processors, high-performance interconnects, and large memory footprint all within a single rack space. The A+ Ultra platform’s storage flexibility and performance are perfect for computationally demanding workloads such as running complex simulations or research projects to address these scientific challenges.

Mammoth was purchased through MNJ Technologies (an Illinois-based information technology company) that served as the logistics and sourcing partner for the Supermicro HPC solution.

Paul Kozak, MNJ Technologies COO, stated that MNJ Technologies and its employees are thrilled to make a difference in the country’s health and well-being. MNJ Technologies COO Paul Kozak said, “This gives me great pride knowing that I played a small role in stopping the spread of the virus that has so severely affected so many.”

For more, visit: https://hpc.llnl.gov/hardware/platforms/mammoth

Advanced Micro Devices, Inc. trademarks AMD, EPYC, AMD logo, and combinations thereof

Read the original article here.