People to Watch 2022

Katherine Riley
ALCF Director of Science, Argonne National Lab

Katherine, congratulations on being named a 2022 HPCwire Person to Watch! Can you give us a summary overview of your responsibilities at the Argonne Leadership Computing Facility and what your position entails?
Thank you. It’s an honor to receive this recognition. As Director of Science at the Argonne Leadership Computing Facility, I help oversee the ALCF’s science mission to provide supercomputing resources that enable researchers to achieve breakthroughs in science and engineering. I also help to chart future scientific directions for the facility (more on that later!). This past year, I also took the reins as the project manager for the INCITE program, the flagship allocation program for DOE’s leadership computing facilities.

You’ve been active on the event circuit, discussing how Argonne is contributing to the Exascale Computing Project and preparing to host an exascale supercomputer, Aurora. Please highlight some of the successes that Argonne has had on the path to exascale. (HW, SW, applications, people – anything!)
One of the things I most excited about is all the work that’s being done to develop a portable ecosystem for exascale computing. The Argonne team, along with many collaborators in the Exascale Computing Project (ECP), have really been pushing to make several programming models available across all of DOE’s exascale systems. This work is all aimed at creating an environment that can help reduce some of the burden of developing applications in a more diverse accelerator universe. We’ve had a single accelerator universe for a long time but we’re clearing expanding. We want to create an environment where scientists don’t feel locked in to any one technology and can explore using new and different hardware.

I’ll also point to training as an early success. In addition to conducting Aurora workshops and hackathons for teams in ECP and the ALCF’s Early Science Program, we’ve also made it a priority to host public webinars and training series that are designed to educate the broader HPC community on some of the tools and technologies that will help with exascale application development. We’re building an impressive repository of training materials that will be useful to researchers preparing to use Aurora.

Milestones are inspiring and exciting. What excites you most about entering the exascale era? What are some examples of the science and hopefully breakthroughs that will be unlocked? In what ways will having exascale systems – and I mean the entire ecosystem not just the hardware – be game-changing?
It’s an exciting time for scientific computing. We’re seeing is a lot of growth in the complexity of the science that people want to do and that will be reflected across the entire exascale ecosystem. At ALCF, we’ve been supporting more and more projects that are using our systems for things like exploring and integrating learning methods, tackling data-intensive challenges at greater scales than ever before, and tightly coupling experimental work with large-scale simulations. So with the coming exascale supercomputers, we’re not just building big compute systems; we’re also building a more adaptable software ecosystem that will support a wider range of science.

The CANDLE project, which is focused on cancer research, is a powerful example of a science application that stands to benefit greatly from exascale systems. With the increased power and capabilities of these systems, the team is looking to automate the analysis of millions of cancer records to arrive at data- and simulation-informed strategies for personalized treatments. That’s something that is not possible with today’s supercomputers.

ALCF hosts an impressive AI testbed. Why is this hardware important? What is the opportunity for Argonne, for science?
We’re really excited about the AI Testbed. This is one of the new scientific directions were charting at ALCF. We’re working with several AI startups to assemble an array of accelerators to explore the how the technology can be used for next-generation scientific computing and machine learning. Can it be effective? Can we improve science capabilities by collaborating with vendors? Can it be integrated into an HPC environment effectively? The idea behind the testbed is to have the research community play an active role in determining how AI accelerators can be used for science.

Here at Argonne, we’ve already seen some success in using these platforms for things like COVID-19 studies and tokamak research, but we’re looking forward to opening the testbed resources up to the larger community this year. We’ll be accepting project proposals and hosting some training workshops very soon, so stay tuned!

Heterogeneous computing architecture, largely dependent upon accelerators (GPUs mostly), has become the dominant approach to supercomputing (with the notable exception of Top500 leader Fugaku) and is the backbone of the U.S. exascale program. Where do you see computer architecture headed? What will be the follow-on to today’s dominant heterogeneous (CPU plus accelerator) landscape?
We are definitely in an accelerator-driven space. The exascale systems coming online soon will have just one type of accelerator per system. Further down the road, I think we’re likely looking at a much broader range of heterogeneity with multiple different types of accelerators. But we’re not yet sure what this will look like. Will we have configurable nodes? Can different accelerators be leveraged for different purposes? Will that be too expensive? These are some of the questions we’re exploring with the ALCF AI Testbed.

Are there any other computing trends you would like to comment on? Any areas you are concerned about, or identify as in need of more attention/investment?
Scientific application and software sustainability are absolutely crucial as we continue to move to a more heterogenous computing world. There are many efforts going on in this space, but more is needed. ECP, for example, has done an incredible job building a software environment for a large set of important scientific applications. The ongoing work to develop portable programming models and libraries will benefit an even larger set of applications. But the world of scientific computing reaches far beyond that and developing sustainable applications is a big lift that requires more investment, training, and people power.

What inspired you to pursue a career in STEM and what are your suggestions for engaging the next generation of domain and computational scientists?
I knew I wanted to study science since around fourth grade. One of my favorite books was “A Wrinkle in Time,” where the father was an astrophysicist. Then I spent many years learning what an astrophysicist really does. I didn’t end up studying astrophysics but became very excited by HPC research. How do we write good and performant scientific applications? I had a strong stubborn streak and perseverance that kept me going, but exposing the next generation to science and life as a scientist is so important. Some of the outreach from Argonne, like Introduce a Girl to Engineering Day and our summer coding camps, reach school-age kids and give them an opportunity them to interact with our scientists, showing them what’s it like to work in STEM and hopefully sparking their interest in science careers down the road. I think this type of interactive, hands-on outreach is critical to reducing the mystery of what scientists do and showing them that science is for everyone.

Outside of the professional sphere, what can you tell us about yourself – unique hobbies, favorite places, etc.? Is there anything about you your colleagues might be surprised to learn?
One unique hobby is Olympic-style weightlifting. I was working to rehab a really bad knee, saw some people doing it, thought it was interesting, and started pursuing it for fun. It has become my favorite form of recreation that is not outside.

Another big hobby is hiking. But recently, the focus has been more on working to get my young children to enjoy hiking with us!

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

HPCwire