High-Performance Computing

U Michigan Mixes Physics Modeling with Big Data

• By Dian Schaffhauser
• 09/22/15

The University of Michigan will be mixing supercomputer simulations with big data. A new high performance computing instrument named "ConFlux" is under development, to be outfitted with next-generation central and graphics processing units, large memories and ultra-fast interconnects to handle data-intensive operations for physics-oriented predictive modeling.

That work is being funded by $2.42 million from the National Science Foundation as well as $1.04 million from the university. The resource will reside in a new Center for Data-Driven Computational Physics within the Michigan Institute for Computational Discovery & Engineering.

The expectation is that the new computing resource will suit the kind of physics-oriented simulations needed for advances in five areas: aerodynamics, climate science, cosmology, materials science and cardiovascular research. What they have in common is the complexity of atoms and molecules. As the researchers wrote in their NSF proposal, "a wide range of phenomena exhibit emergent behavior that makes modeling very challenging."

Frequently, calculations must resort to approximation, said Karthik Duraisamy, an assistant professor of aerospace engineering and director of the new center, in a prepared statement. "Such a disparity of scales exists in many problems of interest to scientists and engineers."

For example, Duraisamy will lead a project to create a turbulence simulation. The goal will be to more accurately model the flow of air or water as it breaks up into swirls and eddies to develop more efficient airplane designs, improve weather forecasting and address other fields that involve the flow of liquids or gases.

Course material developed from the usage of ConFlux will be added to the curriculum of several computational and data sciences degree and certificate programs. Also, use of the ConFlux cluster will be extended to research groups outside of U Michigan through the use of Extreme Science and Engineering Discovery Environment (XSEDE), an NSF-funded project that lets scientists share computing resources, data and expertise.

ConFlux will also specifically be made available as a resource to minority-serving institutions and historically black colleges and universities. Eventually, middle and high school students will also gain access through educational modules in computing, physics and data developed by the Society of Women Engineers at U Michigan.