Next-generation supercomputing capability promises major discoveries, scientific advancements, and new applications in research
By Rob Johnson
“Exascale systems are the fastest computers humans have ever built. Because of that, we will soon attack extremely complex problems not possible before,” said Rick Stevens, associate laboratory director for computing, environment and life sciences at Argonne National Laboratory and professor of computer science at the University of Chicago. As a central figure behind Argonne’s forthcoming Aurora exascale system, Stevens understands intimately the major impact exascale computing will have on the world as we know it. Aurora is the front runner as the first exascale system deployment in the United States, and possibly, anywhere around the globe.
Once online, Aurora is expected to exceed the world’s most powerful high-performance computing (HPC) systems by a large margin. With enthusiastic optimism, Stevens describes the raw computing power of the forthcoming system. “For the first time, we will have computing capability capable of ten to the eighteenth power operations per second.” In other words, Aurora will achieve performance in the exaflop range of a billion-billion calculations per second.
“We chose the name Aurora to represent our aspirational goal to create something which in some sense can illuminate the world,” said Stevens. “Exascale computing will help us gain greater understanding of global climate, assist in the design of medical treatments for diseases, design new energy sources, understand the nuanced inner workings of atoms, simulate the neural connections of something as complex as a brain, and so much more.”
In 1946, Argonne was chartered to develop nuclear reactors for the nation’s peaceful nuclear energy program. Since that time, the laboratory’s mission has increased in scope dramatically. Today, the team at Argonne, along with researchers around the globe, spearhead research in diverse scientific disciplines including physics, chemistry, material science, computing, biology, and genomics. Argonne supports about 3,500 researchers with a billion-dollar budget each year. Added Stevens, “Unlike many university research environments which face budget and staffing constraints, Argonne’s resources make it possible for us to dedicate larger teams and technologies to tackle extremely difficult, long-term scientific problems. As a result, we can support breakthrough science every day.”
Integrating simulation, data, AI, and machine learning
What makes exascale computing so special? Aurora will further advance and accelerate scientific research and discovery by harnessing the convergence of high-performance computing and artificial intelligence (AI). The exascale supercomputer’s revolutionary architecture will be capable of supporting AI, machine learning, and data science workloads alongside traditional modeling and simulation research, enabling scientists to tackle larger and more complex problems.
“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,” said Stevens. “This idea of mixing simulation and data-intensive science will give us an unprecedented capability, and open doors in research which were inaccessible before, like cancer research, materials science, climate science, and cosmology.”
One example of this combined approach is “training” autonomous vehicles. One training method involves video capture to help the self-driving technology understand basic road rules and conditions under common driving circumstances. However, for self-driving cars, the unexpected and potentially hazardous scenarios – like a person walking into traffic – are far more challenging for a system to anticipate. To mitigate that risk, simulations of unusual accident scenarios help the car learn to interpret potentially dangerous scenarios and choose the safest action. The combination of approaches makes it possible to develop autonomous driving technology to cover all ends of the behavioral spectrum.
With Aurora, Stevens foresees the incredible potential for science. “For the first time, we can bring the best simulations and the highest quality observational data together to build models of the world in ways we couldn’t do before. That’s why exascale systems will change pretty much everything.”
This approach has potential applications in many other fields such as cancer prevention and treatment. “For some common types of cancer, we have a lot of data,” described Stevens. “However, there are also very rare cancers. We need to find a way to use our understanding of biology to create more data regarding those rare cancers. In doing so, we will gain much better insight as to how they respond to drugs and find ideal treatments. In medical applications like this, we foresee exascale computing quite literally saving lives.”
Exascale: A team effort
“Many people have contributed to the success of Aurora. The team has been working hard for many years to design the system, optimize it, and to create the specialized innovations in software and hardware necessary to make this work,” said Stevens. “It is a fascinating technical challenge which caused all of us to think differently about ways to design computers, how to make them more reliable, and more energy efficient. If we can improve energy efficiency in these very large-scale systems, that benefit will trickle down for improved energy efficiency in cellphones, laptops, tablets, electric cars, and other technologies.”
“Building the country’s first exascale machine is a very complex endeavor, and we feel very fortunate to have two wonderful partners – Intel and Cray – who are committed to the success of Aurora. Intel brings to the table both their technological depth of experience and highly performant hardware such as future Intel Xeon Scalable processors underlying Aurora. Without advanced hardware, there’s no way we could push the exascale computing envelope. Cray supports us through the software overlay which makes the integrated system possible. Together, we are making a new era of exascale computing possible. It’s an exciting time.”
Of course, an effective and efficient system needs optimized applications too. The nation’s investment in the U.S. Department of Energy’s (DOE) Exascale Computing Project (ECP) has resulted in the Extreme-Scale Scientific Software Stack (E4S) release which represents a major step forward. The ECP is also working to develop over 25 applications designed to function at the exascale level. In addition, the Argonne Leadership Computing Facility’s Aurora Early Science Program (ESP) is supporting multiple research teams’ efforts to prepare key applications for the scale and architecture of the future exascale system. “At day one, we’re going to have a large number of applications, not just one or two but 20 to 30 applications ready to go,” said Stevens.
Ready, set, go!
Once online, Aurora promises to make short work of some very complex problems. With so many potential applications, what will Aurora tackle first?
According to Stevens, many researchers and scientists will vie for compute cycles on Aurora. “We have over a thousand people working on building application tools, and they eagerly await access to these machines. Interested researchers seek breakthroughs for materials science, fundamental physics, more efficient wind farms, building safer nuclear reactors, and so much more. All of these topics are pursued now, and with great enthusiasm.”
Stevens adds, “When exascale machines first turn on, we anticipate it’s going to be a crowded dance floor. Exascale computing is an incredibly powerful tool which will help us in almost unimaginable ways. Our team is proud to help lead the way into next-generation computing.”
Rob Johnson spent much of his professional career consulting for a Fortune 25 technology company. Currently, Rob owns Fine Tuning, LLC, a strategic marketing and communications consulting company based in Portland, Oregon. As a technology, audio, and gadget enthusiast his entire life, Rob also writes for TONEAudio Magazine, reviewing high-end home audio equipment.
# # #
This article was produced as part of Intel’s HPC editorial program, with the goal of highlighting cutting-edge science, research and innovation driven by the HPC community through advanced technology. The publisher of the content has final editing rights and determines what articles are published.