The Department of Energy (DOE) has widely promoted the roll out of three Exascale systems over the next three years, all with HPE Cray supercomputers. The Aurora system at Argonne National Laboratory and Argonne Leadership Computing Facility (ALCF), Frontier at Oak Ridge National Laboratory Leadership Computing Facility (OLCF) and El Capitan at Lawrence Livermore National Laboratory (LLNL) mark the next milestone in high-performance computing (HPC).
Imagine a computer so powerful that it can predict future climate patterns, saving millions of people from drought, flood, and devastation. Or a computer so powerful that it can model a cancer cell with such accuracy that we can create a personalized treatment, just for you.”
Supercomputing performance is measured in FLOPS (floating point operations per second). The world’s fastest supercomputers today can reach hundreds of Petaflops—or 10^15 floating point operations per second. And by next year, the world’s fastest supercomputers will top an Exaflops.
Exascale supercomputers will perform a quintillion calculations every second. That’s 10^18 FLOPs—a billion billion or 1,000,000,000,000,000,000. To put that number in perspective, it would require every person on Earth calculating 24 hours a day for over four years to do what an Exascale supercomputer can do in 1 second.
The Exascale Era will usher in enormous opportunities, from breakthroughs in drug research (cancer and COVID-19) to developing safer, more efficient vehicles, making clean energy advancements and understanding our climate crisis. Such technological advances will help corporations and startups analyze massive data sets and improve AI and Machine Learning.
Rick Stevens, the Associate Director of Computing, Environment and Life Sciences at Argonne added “What excites me most about Exascale systems like Aurora is the fact that we now have, in one platform and one environment, the ability to mix simulation and artificial intelligence. This idea of mixing simulation and data-intensive science will give us unprecedented capability, and open doors in research which were inaccessible before, like cancer research, material science, climate science and cosmology.”
Exascale Day 10.18
Because Exascale computers can compute 10 to the 18th calculations each second, October 18 (10/18) has been declared National Exascale Day. It’s the official day of recognition for the supercomputing and broader HPC communities and the role they play in our world.
Argonne National Laboratory Director Paul Kerns captured the spirit of Exascale Day, “Imagine a computer so powerful that it can predict future climate patterns, saving millions of people from drought, flood, and devastation. Or a computer so powerful that it can model a cancer cell with such accuracy that we can create a personalized treatment, just for you.”
At Argonne, researchers will be able to use simulation, data-intensive, and learning applications (machine and deep learning) to drive Exascale-speed research in the fields of fusion reactor plasmas, computational fluid dynamics, neuroscience and drug response prediction. Using Frontier, OLCF scientists will accelerate science and technology innovation in fields such as fusion energy, disease genetics and materials science. And at LLNL, the US National Nuclear Security Administration will use El Capitan to improve the science and technology surrounding the maintenance and modernization of the US nuclear arsenal.
Today, global digital transformation and the urgent need to extract insight from rapidly growing data has prompted a major shift in computing requirements. Traditionally siloed workloads such as analytics, AI, simulation and modeling and IoT are fusing into one business critical workflow operation at unprecedented scale. These requirements are driving the rapid evolution of supercomputing infrastructure. At the same time, they are enabling the broader adoption of supercomputing technologies to power innovation and digital transformation across every research and commercial data center.
Dr. Robert W. Wisniewski, Intel’s Chief Architect HPC and the Aurora Technical Lead, discussed how Intel’s partnership with Argonne and the DOE will impact the future. “Supercomputers of the Exascale Era are going to allow new discoveries in science and engineering and accelerate advances that will benefit and propel humanity forward. Achieving Exascale and beyond will require leveraging heterogeneity. We at Intel are delighted to be partners with Argonne and the broader DOE as we deliver Aurora and drive the technology to meet their mission needs.”
Delays for Aurora announced
The DOE believes that the United States will stand up an Exascale system in 2021 as promised. The bad news is that Aurora will not be the first American Exascale-class supercomputer, as originally planned. Currently it looks like Frontier (OLCF) will be the first Exascale system stood up.
“At the inception of the Exascale Computing Project (ECP) we asked researchers to imagine new frontiers in science and engineering enabled by Exascale computing. With Frontier, we have the opportunity now to fully realize our original vision, solving grand challenge problems that lead to breakthroughs in areas of energy generation, material design, earth and space sciences, and related fields of physics and engineering,” said Andrew Siegel of Argonne.
DOE Under Secretary for Science Paul Dabbar announced that the first US Exascale system will not, as promoted for years, be the Intel-powered Aurora system at Argonne. It will instead be HPE-Cray’s Frontier, powered by AMD CPUs and GPUs and designated for Oak Ridge National Laboratory. Innovation DuPage has spent it’s first three years promoting the launch of Aurora, since receiving a thorough introduction to the true capabilities of the system at The Center for Efficient Exascale Discretizations (CEED) a co-design center within the DOE Exascale Computing Project (ECP) and at SC18 and SC19 - the international supercomputer conference. Innovation DuPage will continue to track and promote Aurora’s progress, impact on AI, the startup community and economic development opportunities throughout the DuPage region.
Citing the strategy of DOE’s Exascale Computing Project (ECP), Dabbar said the first three Exaclass systems will be complemented by an applications and development ecosystem that has been under construction for several years.
“A big part of the deployment … is really about the whole system,” Dabbar stated. “The software stack that is going to be riding on top of the hardware is going to be integrated. Part of the deployment is the layering of the operating system on top of the hardware.”
The most impactful example for ID is SPACK (spack.io), the Supercomputer PACKage manager which was a R&D 100 Award winner because of its worldwide impact on HPC. Todd Gamblin, the creator of SPACK, described it as the Flexible Package Manager that automates the software on the Exascale supercomputers that simplifies the building, installing, customizing and sharing of HPC Software stacks. Spack’s original 100 software packages have grown into a library of more than 4,500, with an active open source community of more than 3,000 users and 700 contributors regularly adding features and improvements.
Spack is now used for software deployment on six of the world’s top 10 supercomputers and will be utilized on Aurora, Frontier and El Capitan.
The Exascale Computing Project (ECP) offers a collection of reports encompassing the software stack, application development and hardware. Please review the ECP Exascale Day collection which includes videos, podcasts and research at https://www.exascaleproject.org
This is the first in a series on the growth of Deep Tech development, emerging startups and venture investment that will impact the DuPage region.
Travis Linderman has served as director for three venture incubators prior to his selection as Managing Director of Innovation DuPage. He founded a startup with Princeton University, secured venture capital backing, scaled rapidly and enjoyed a successful IP acquisition. Travis has spearheaded capital campaigns that have raised over $700 million for technology development. Travis serves as the MD of Exascale Ventures.
Quotes and source materials provided in part by interviews by Inside HPC (Doug Black) and ECP