Boston University engineering and science researchers now have top-tier computing power that enables them to run simulations of unprecedented complexity on campus. A gift of hardware from Hewlett-Packard Co. now enables researchers to create simulations in three dimensions as they explore disciplines like aerodynamics, bioinformatics, particle physics and others.
The gift of 160 graphics processing units – or GPUs – increases the power of an existing system fivefold to 80 teraflops in peak performance. GPUs have a capacity that is useful for researchers seeking mathematical representations of highly complex models. Previously, if researchers wanted to run these kinds of complex simulations, they needed to apply for time on the massive computers housed at government-funded national laboratories around the county, a highly competitive and time-consuming process.
Bryan Marler, Hewlett-Packard’s director of high performance computing for R1 universities and research labs, said the new BU system “is the most powerful GPU system in academia on the East Coast and one of the top 10 in the country.”
Assistant Professor Lorena Barba (ME) is one of the BU researchers who welcomes the additional power of the new system. She is working with a biologist at Virginia Tech studying a species of Southeast Asian snake that glides in the air between trees. The biologists have video of the snake in flight, but that doesn’t provide the engineer with enough understanding of the aerodynamics the snake uses to maneuver in air.
“These snakes are very good gliders, but you can’t put a snake in a wind tunnel,” Barba said, so computer modeling is the most effective way for the engineer to study it. Barba, who has won several honors for her GPU-related research, hopes to glean enough data from the video and other sources, feed them into the computer and emerge with a three-dimensional model of the aerodynamic forces at work.
“With the old system, we could do some smaller scale modeling, but only in two dimensions,” Barba said. “It was limited. Now, with the new system, we have an exponential increase in capability and we can model them in 3D.”
The first system, established in 2009, quickly found users across departments in ENG and CAS. Original users Professor Richard Brower (ECE), Professor Claudio Rebbi (Physics) and Barba reached out to other computing-intensive research groups on campus and demand for the GPU system spiked. The expanded GPU system addresses the increased demand.
In addition to the College of Engineering and the College of Arts & Sciences, the Center for Computational Science and the Office of Information Services & Technology contributed to the GPU expansion project.