Randles Selected to Help Pilot First U.S. Exascale Computer

Amanda Randles, the Alfred Winborne Mordecai and Victoria Stover Mordecai Assistant Professor of Biomedical Engineering at Duke University, is getting an early crack at using Aurora—the nation’s newest and largest supercomputer.

Housed at Argonne National Laboratory (ANL), Aurora will be the nation’s first exascale supercomputer, capable of performing a quintillion calculations per second upon its expected 2021 completion. This is more than five times the maximum theoretical limit of the world’s current faster supercomputer Summit that debuted earlier this year. The Aurora Early Science Program (ESP) selected 10 simulation-based projects to help prepare key applications, libraries and infrastructure for the architecture and scale of the exascale supercomputer.

Randles is one of five researchers outside of ANL selected to participate. She will be leveraging Aurora’s record-breaking power to improve HARVEY—a code capable of simulating blood flow through the human vasculature down to the cellular level.

“HARVEY should run on Aurora without any difficulties,” said Randles, who has been developing HARVEY for more than half of a decade. “And this allocation will allow us to understand how to best leverage this next generation hardware and hopefully achieve unprecedented scale with our code.”

The goal of these simulations is to develop a better understanding of the role biological parameters play in determining where tumor cells circulating in the bloodstream end up. Randles eventually hopes to develop HARVEY to the point that clinicians can use it in real-time to determine a specific individual’s risk of metastasis as well as other health complications.

The added computing power from Aurora will allow the program to simulate larger regions of the body for longer periods of time at higher resolutions. And with the added computing power comes funding for a postdoctoral researcher at ANL to work full-time on preparing HARVEY for the transition. The added help will allow Randles to develop new methods for in situ visualization and analytics.

“There will be so much memory used for the simulation that it’s really not tractable to output all of the data for post-processing,” said Randles. “Being able to visualize the simulation while it’s happening and analyze results dynamically will allow us to use the large amount of data to its fullest.”