High Performance Computing Accelerates Discovery
NADIA M. WHITEHEAD | August 29, 2013 | UNIVERSITY COMMUNICATIONS
Ricardo Bernal, Ph.D., left, and Sudheer Molugu, right, use the
University's high-performance computing system and a $2 million
cryo-electron microscope to conduct research on chaperonins,
Photo by Nadia M. Whitehead / UTEP News Service
Not long ago, one computer was enough to solve anyone's problems. But these days, hundreds, or even thousands of computers working together may be a researcher's best bet at success.
"Some of the problems people are solving these days are so big that they need high-performance computing power," said Pat Teller, Ph.D., director of research computing at The University of Texas at El Paso. "And if you are using 100 processors rather than just one, you might be able to execute your large problems in 1/100 of the time."
Teller is responsible for the new Research Cloud @ UTEP, which is housed in the brand new Research and Academic Data Center (RADC). The Research Cloud provides users with up to 31 high-performance servers – almost 400 processors – that are connected via a 10-gigabyte network. Even bigger systems are available to UTEP researchers through The University of Texas Research Cyberinfrastructure (UTRC) initiative and Extreme Science and Engineering Discovery Environment (XSEDE), funded by the National Science Foundation.
Users of UTEP's high performance computing (HPC) system can be found in nearly every field – from mathematicians and physicists to chemists and mechanical engineers.
"The possibilities are endless with HPC," Teller said. "You can predict and model the weather, sift through the huge digital collections of the Library of Congress to find what you're looking for, and model and simulate the strength of materials."
For instance, Ricardo Bernal, Ph.D., associate professor of biochemistry, is using UTEP's HPC system to create three-dimensional reconstructions of chaperonins, or proteins.
Using a $2 million microscope known as the cryo-electron microscope (cryo-EM), Bernal and Sudheer Molugu, the cyro-EM facility manager, freeze the proteins and take thousands of photos of them under the microscope.
"The proteins are frozen into the ice in many random orientations – so that's where high performance computing comes in," Bernal said. It's the computer's job to determine the orientation of every protein relative to every other protein. The computer then averages the thousands of images together to ultimately figure out what the 3-D structure looks like.
If the team can figure out exactly what the protein structure looks like and how it works, they can they apply the information to address diseases associated with chaperonins.
Without HPC, Bernal and Molugu's research would slow down excessively, or even come to a standstill.
"Having the UTEP high performance computing resources available to us has been absolutely critical to advance our research," Bernal said. "We are able to speed up computational analysis of the cryo-EM data by at least tenfold when compared to a powerful lab workstation."
Carl Dirk, Ph.D., a UTEP professor of chemistry who studies molecular reactions, only recently discovered HPC at UTEP, and is beyond excited about its capabilities.
For most of his life, he has used semi-empirical calculations, which are approximations, and no longer considered "cutting-edge" math.
"In my opinion, it's a good, predictive method still, but a lot of journals no longer accept semi-empirical results – only those using ab initio methods, which I did not have the ability to perform," he said.
In July he started performing ab initio calculations on the system. He expected them to take weeks to complete – they were finished in less than 12 hours.
"This new computing system has made me rethink what I can do in terms of computational chemistry. It's changed my mind about what I thought we could do at UTEP," he said. "Up until now I thought this would be very difficult to do – that I could never be competitive with other top researchers, but now I can be."
There are currently about 35 users of UTEP's HPC system, but Teller hopes that more University faculty, staff and students will hop on the bandwagon soon to learn what HPC can do for them.
She said, "If people engage themselves and adopt this computing, it can certainly accelerate their research, time-to-results, and even grant funding."
For those interested in using the system, Research Computing @ UTEP offers one-day and three-to-five-day training courses throughout the year. Learn more at researchcomputing.utep.edu.