As high performance computing creates new possibilities and opportunities for research, it also generates an ever-increasing need for improved performance capabilities of the machines themselves.

While a person using their own computer might have several applications open at the same time and not experience a loss of capabilities, when researchers are working on the cutting edge of technological revolution, it is often imperative that what they are doing in the moment isn’t at the expense of other programs and vice versa.

Michela Taufer

Now, a team including Michela Taufer—UT’s Dongarra Professor in the Min H. Kao Department of Electrical Engineering and Computer Science—and researchers at Lawrence Livermore National Laboratory (LLNL) have proposed a possible solution that was recently acknowledged with a prestigious R&D 100 award: Flux.

With Flux, the idea is that programming that might otherwise cause a drain on processing speed would instead be scheduled for particular times, helping make certain that computing operations could take place when it would result in optimal performance overall.

“Instead of the traditional method of running programs or software when users initiate them, our system figures out what the most desirable workload is to make operations work the fastest and the smoothest, and how to schedule it,” said Taufer. “By helping solve performance issues, it not only makes large-scale projects or research that depends on high-performance faster and better, but it results in reduced energy needs, which can really make a difference over time.”

In addition to Taufer, the team behind Flux includes LLNL computer scientists Dong Ahn, Albert Chu, Jim Garlick, Mark Grondona, Stephen Herbein, Daniel Milroy, Christopher Moussa, Tapasya Patki, Thomas Scogland, and Becky Springmeyer.

Together, they worked to give Flux an open-source framework that could be adapted to a wide range of workflows on any number of computing interfaces, including cloud-based, on next-generation architectures, from remote locations, and even from laptops, in addition to the obvious role it can play for massive supercomputers and in exascale computing.

Because of the adaptive nature of Flux, potential areas of impact are as varied as the aforementioned types of systems where it can be used, with COVID-19 modeling, cancer research and drug design, engineering and design optimization, and artificial intelligence all being areas mentioned where it could be a game-changing technology.

“The key to Flux is that it allows for a much more robust control of computational workflows, giving researchers the ability to solve critical problems facing society,” said Taufer.

The awards are among highest accolades in technology and engineering, having drawn comparisons as the scientific equivalent of entertainment awards like the Oscars or Golden Globes. Nominees come from 17 different countries or regions around the world, with a round of finalists selected about a month before winners are revealed.