When Montana State University researcher Mark Owkes needs to predict the movements of millions of tiny droplets of fuel and molecules of air inside a vehicle’s combustion chamber, he needs some serious computing power.
Up until this year, Owkes, an assistant professor of mechanical engineering, probably would have had to do those computations somewhere other than MSU.
But this year, the university launched an initiative to put high-speed, high-powered computing resources into the hands of every MSU researcher who needs them – without having to leave campus.
“No one thinks of us in this space right now: high-performance computing,” said MSU’s Chief Information Officer Jerry Sheehan.
But that was before MSU’s Hyalite Cluster, an interconnected bank of computers in Bozeman that combines to form a supercomputer.
Operated and maintained by MSU’s Research Cyberinfrastructure group, the cluster has about 1,400 computer cores, Sheehan said. Altogether, it has roughly 2,800 gigabytes of RAM and almost 900 terabytes of storage space.
“The Hyalite Cluster was made possible by working with faculty to pool their resources with MSU’s Information Technology Center to create something bigger than any of us could have managed on our own,” Sheehan said. “It’s a great collaboration.”
On top of that, some of the older, smaller clusters that existed around campus before Hyalite have been repurposed, creating discovery environments for undergraduates.
“It gives them access to technology to make sense of the raw material that is data,” Sheehan said. “What we’re trying to do is give faculty and student researchers the resources they need to be creative and focus on the science as opposed to worrying about the infrastructure.”
Owkes, who joined the faculty in 2014, has been using large computer clusters his entire career. This year, he was awarded a three-year, $276,365 grant from the National Science Foundation. His research focuses on gas and liquid flows, such as the fuel spray in a car’s fuel injection system. It could lead to more fuel-efficient engines that produce less pollution.
“We know how combustion works, and we can model that. But modeling a spray turns out to be very challenging,” he said.
Owkes’ simulations require large amounts of computational resources. Just the different scales between a simulated engine and the tiny fuel droplets, only a few millionths of a meter across, makes the problem impossible to handle on a desktop computer, he said.
Trying to put that research onto a big, government-run cluster requires a bureaucratic process that involves drafting a formal proposal that is reviewed by a scientific review group taking months. If you are awarded cycles on these resources you then jump into the national que where it may take days before your job starts, Sheehan said
“You end up wasting so much time you can’t really accomplish anything,” Owkes said. “Having a medium-sized cluster here on campus makes research possible.”
While it’s true that Hyalite’s not beefy enough rank anywhere near the top 500 supercomputers in the country, Sheehan said speed isn’t so much the point as addressing the computing needs the university actually has.
Sheehan noted that almost two dozen labs and more than 40 users across campus, from neuroscience to chemistry to biology have made use of Hyalite already.
Contact: Jerry Sheehan, (406) 994-2525, email@example.com.