The next scientists | News | National Center for Supercomputing Applications at the University of Illinois
The next scientists
11.09.10 - Permalink
by J. William Bell
Undergraduate institute prepares students to integrate computer science, other research disciplines
Integrated learning is the driving principle behind the Undergraduate Petascale Institute, which is training a new generation of scientists. They'll be computational scientists whofrom the very beginning of their education and careersmake computational simulation an integral part of their research and understand how those simulations can drive discovery.
The institutea collaboration between NCSA and Shodoris part of the Blue Waters project. It kicked off this summer, with more than 20 students giving up their Memorial Day weekend, then spending the next two weeks at NCSA in an intensive computer science workshop. At the workshop, they learned about the architecture of the Blue Waters supercomputer and the programming languages they will need to use the computer. They also found the time for some Ultimate Frisbee.
These skillswell, maybe not the Frisbeewill be applied more broadly throughout the students' careers.
"The program shows the different methods that are currently available, and then it gives you enough of an idea of the way that this thought process works about how to keep a lot of different cores busy and how to chug through your data as fast as you can. But, if you have that general framework, you can come up with newer, better ways to do this," explains Brandon Holt, an undergraduate from the University of Wisconsin, Eau Claire, who took part in the institute.
What is a computational scientist?
"The mind is very powerful, but it's only so powerful. And it's very creative, but it doesn't have a lot of computing power. So I kinda look at computational science as sort of an add-on to the human brain, just giving it more power," says Stockton College's Michael Laielli, another participant. "I think the computational scientist is a translator between the real science and the computer science."
Part of that translation work is simply embracing the world as it is instead of trying to shoe-horn it into hidebound computer science standards.
"All the stars interact at the same time. All the electrons in an atom interact at the same time," says Shodor's Bob Panoff. "If we're going to get better science off these machines, especially a machine like Blue Waters, which is going to have hundreds of thousands of processing cores, we're going to have to be able to map the parallelism in nature more closely to what the machine is capable of computing."
'An apprentice researcher'
During the institute, students learned from experts in the field. And from each other.
"We have a term called 'shoulder surfers.' Whenever we have a lab project or anything in class that they're trying to teach us, they come up behind us and they tap us on the shoulder, and they will sit there. I know plenty of times...we've gone through a project or a lab project with each other, and we've completed it," says North Carolina A&T State University's Brittany Hourston.
Laielli agrees: "A lot of problems, I get answered just by asking the guy next to me, and we figure it out together. That's been a great approach to learning."
The program didn't stop with the summer workshop. Students are paired with faculty mentors at their home institutions.
"What's really important is that students have a chance not only to learn this as content but in the context of a real project with a professor as part of their research group," says Panoff. "They're really becoming an apprentice researcher to a scientist who is passionate and dynamically involved in the conduct of that research."
Together, the institute students and mentors are working on real-world research projects that require computer simulation in fields like chemistry, biology, and astrophysics. Holt, for example, is working on a computing code that does mantle convection modeling of the Earth's core, exploring the use of graphics processing units to speed up the computation and look at the process in finer detail.
"The only way we're going to succeed is to get students into a mode where it's not simply coming up with an answer, but it's thinking about how they came up with an answer," Panoff says. "They'll be able to have a better sense of not just solving the problem right but solving the right problem."