Frontier is an exascale computer planned for delivery at the Oak Ridge Leadership Computing Facility in 2021. The system will support a wide range of scientific applications for advanced modeling and simulation, as well as high-performance data analytics and artificial intelligence. In the “Science at Exascale” Q&A series, researchers working on these next-generation scientific applications discuss what they hope to achieve on Frontier.
Led by the US Department of Energy’s (DOE’s) Lawrence Berkeley National Laboratory, the ExaStar project is working to deliver an exascale application for stellar explosions, including supernovae and neutron star mergers, as part of the DOE’s Exascale Computing Project (ECP). Such simulations will help scientists explain the origin of elements heavier than iron in the universe and answer other fundamental physics questions, such as the source of gravitational waves and the behavior of matter at extreme densities.
The results from ExaStar simulations may not only lead to new discoveries through computation but provide important information to guide US nuclear science experiments.
In this interview, Bronson Messer, a senior scientist at DOE’s Oak Ridge National Laboratory and member of the ExaStar team as part of the ECP project, describes what researchers expect to achieve on exascale systems like Frontier.
What are some of the big science questions that the ExaStar application will help address?
Messer: With ExaStar, we’re trying to answer a variety of questions in nuclear astrophysics. Why is there more iron than gold in the universe? Or, for that matter, why is any element rarer than another? How are space and time warped by gravitational waves? How are neutrinos and subatomic particles produced in stellar explosions? But the challenge to us as scientists and code developers is including so many different pieces of physics—from the small-scale transport of subatomic particles and nucleosynthesis of the elements, all the way to the large-scale hydrodynamic motions of the stars as they merge or explode under the influence of general relativistic gravity, which is the bowl in which this is all boiling. Computing all these processes at high fidelity becomes very complex.
What do you hope to do on Frontier that you can’t today?
Messer: Because we have so many different pieces of physics to contend with, we have many, many variables we must track at each point of the simulation, so the ability to have a large amount of very fast memory like we’re going to have on Frontier will be a real boon to our simulations. With the level of improved speed that we’re going to get on Frontier, we can add more realism and get science results in a shorter amount of time. Instead of taking perhaps on the order of a month to run a supernova or neutron star merger simulation—which is what it takes us today—we may be able to complete this type of simulation in less than a week.
The Frontier exascale supercomputer is slated for delivery in 2021. Until then, what kind of work is the ExaStar team able to accomplish on Frontier’s predecessor at ORNL, the Summit petascale supercomputer?
Messer: The pre-exascale Summit supercomputer has allowed us to take pieces of the microphysics that we need to include in ExaStar simulations and improve the physical fidelity in a lot of ways. For example, we were able to take the nuclear networks—the set of equations that tell us how one element transmutes into another set of elements—and expand a sort of schematic, 13 element network to hundreds of isotopes. With this new level of fidelity, we can actually make predictions that can be matched to telescope observations and push for the next generation of improvements we want to include at exascale.
UT-Battelle LLC manages Oak Ridge National Laboratory for DOE’s Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE’s Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit https://science.energy.gov.