Scientist Is Paving the Way Toward the Next Generation of Supercomputers

As a student who was always drawn to some of science’s most difficult challenges, Kevin Brown found one in the pages of a technical magazine that stuck with him. The story was about building better supercomputers with graphics processing units, or GPUs — the same types of computer chips you’d find in a gaming console.

Brown became fascinated with supercomputing. After earning his bachelor’s degree in computing and information technology at the University of Technology, Jamaica, he went on to complete master’s and doctoral degrees in mathematical and computing sciences at the Tokyo Institute of Technology.

In 2019, he joined the U.S. Department of Energy’s (DOE) Argonne National Laboratory as a postdoctoral appointee in the Argonne Leadership Computing Facility (ALCF), a DOE Office of Science user facility. Two years later, he became Argonne’s first Walter Massey Fellow. The three-year fellowship recognizes an exceptional, early-career scientist committed to diversity, equity and inclusion.   Brown’s work at Argonne to optimize supercomputer performance complements the development of Aurora, the ALCF’s new exascale machine. ​“We’re hoping that the information we get from my studies will tell us the best ways to use Aurora’s network,” he said.

Here, he talks about his research and his experience at Argonne.

Q: What are you working on at Argonne? 

A: My work is concerned with improving the performance of our supercomputers. Supercomputers are used to solve the biggest questions, from the nanoscale to the cosmic scale and everything in between. They do this by performing a massive number of calculations and moving a large amount of data in a given time.

A supercomputer is essentially thousands of smaller computers that are all connected using a superfast network. All of these computers work together on parts of the same problem. Think of simulating the weather in Chicago: You might have different grids in the city that you can process on different computer nodes. Moving that data around requires very fast networks in our supercomputers.

I’m focused on how we can design better networks for the supercomputers that we have — and for supercomputers in the future. For example, what are the sources of network bottlenecks, and how do we change configurations to have less congestion? Or, how do we improve the routing on these networks so that they move data faster? How do we design new types of networks for future supercomputers to make them more capable for the science we will need done?

Q: How do you come up with solutions to these questions? 

A: Usually that involves a process called performance analysis, which helps us understand how much data is being sent over a network, why the data is being sent to different parts of the network, and why communication may be slow with different activities.

We measure all of that and get an understanding of how to improve the network. One way to do that is by designing new routing algorithms. For example, Google Maps will route traffic to make sure it’s spread out on the roads while avoiding congestion. We’re trying to do similar things for our supercomputer networks to route information efficiently.

Another aspect is designing new network architecture, such as how we connect different switches in the network. This is similar to how you might connect different cities with highways so that moving people is as efficient as moving the data.

Q: What drew you to the Walter Massey Fellowship? 

A: Throughout my academic and research life, I’ve enjoyed answering interesting questions. At the same time, I was always very engaged in the community. As a postdoctoral researcher at Argonne, I was president of the lab’s Postdoctoral Society. And in grad school at the Tokyo Institute of Technology, I was president of the International Students Association.

So, I’ve always felt like there are a lot of benefits from having a strong sense of community engagement, which can inform and enhance the research output. I liked that the Walter Massey Fellowship wasn’t just about answering research questions but also seeing how we can work together and build a better community when we’re answering these questions.

Q: Why does your work matter to society? 

A: I’m building research tools for scientists to answer big questions in their domains. So, I get to work with physicists, chemists, material scientists and astrophysicists.

Supercomputers are used to study things that are either too big, such as the expansion of the universe; too small, such as how to layer atoms on top of a surface; too dangerous, such as what happens inside a nuclear reactor; or too expensive, such as whether an airplane can last for 10 years if it is designed a certain way.

We can explore questions like, how should we design infrastructure to be more resilient to climate change? How can we change the human processes that contribute to climate change? These are things we cannot answer using a pen and paper or traditional computers. We need the computational power of supercomputers.

Q: What do you like about being at Argonne? 

A: It is amazing to me that some of the people whose textbooks I’ve been reading for years are now sitting in the office next to me. One of the things I’ve loved about working at Argonne is, the bigger and more difficult the question is, the more excited and undaunted we are as a Lab to try to answer it. I like that drive and passion folks have for solving these big problems.

Q: What advice would you give to someone who wants to follow a similar career path? 

A: First, try to find your network — people who support you in the way that you want to be supported. Once we start talking about our issues and challenges, someone who has been through it may be able to help us. Second, try to be comfortable with the unknown. The only way to reach where we’ve never been is to walk the path that we’ve never walked.