FAYETTEVILLE -- Supercomputers have grown in size and speed and importance, said Jeff Pummill, co-director of the Arkansas High Performance Computing Center at the University of Arkansas at Fayetteville.
"In 10 years time we've seen a profound change in the way our campus does research," Pummill said.
A new acquisition should help UA accommodate increased demand for supercomputing resources, he said, with a supercomputer known as Trestles expected to have users logging on in as soon as two weeks.
"It's very important to us because it approximately doubles our computing capacity," said David Chaffin, the center's other acting co-director.
The university announced acquisition of Trestles in May. The machine had been operating for about four years at the University of California, San Diego.
The name is familiar to surfers, as it's also the designation for a famed surf break near San Diego. The term originates from a bridge in the area.
"I knew Trestles was going to be decommissioned. They were getting a lot larger system," Pummill said. So he asked a friend at San Diego if the computer's host center would consider transferring the machine to UA.
While part of the original Trestles cluster remains in San Diego, now 85 percent of it is at UA, Pummill said.
The university only paid basic expenses to bring Trestles to campus, about $25,000 in moving and installation costs plus $31,000 in warranty costs, according to Chaffin. Pummill said the machinery weighs about 10 tons.
The supercomputer was first built with a $2.8 million grant award from the National Science Foundation announced in 2010. The foundation, which still owned Trestles, had to sign off on the transfer, Chaffin said.
Trestles has 8,192 processing cores. By comparison, a typical laptop has four such processing cores, Pummill said.
In addition to Trestles, the university will continue using a supercomputer known as Razor.
The Trestles acquisition will replace an older supercomputer known as Star of Arkansas that began operating in 2007. In terms of computing speed, the Trestles system is about five times faster than Star of Arkansas, Pummill said.
The Trestles supercomputer operates with a processing speed of 79 teraflops. The Razor unit has a processing speed of 77 teraflops. But Razor has 4,328 processing cores, far fewer than Trestles, which can handle more jobs with more cores.
Other numbers make the acquisition noteworthy. Chaffin estimated that the number of UA supercomputer users has grown more than 25 percent each year for the past four or five years.
About 300 active users now use UA's supercomputers for research projects in diverse fields of science, Pummill said. The Arkansas High Performance Computer Center was officially designated in 2008, Pummill said, but he described its roots as dating back to 2004.
"We are very excited about this," said Salvador Barraza-Lopez, an assistant professor of physics who studies nanoscale devices, which are many times smaller than the width of human hair.
He said he hopes to begin work in the fall to develop a computational tool to help with the engineering of such technology.
"You use the rules of quantum mechanics to design those devices at the atomic level," Barraza-Lopez said, adding that computing resources are very important for such work.
Neither Trestles nor Razor rank among the world's fastest supercomputers, which do calculations at rates more than a hundred times faster than UA's machines.
But Barraza-Lopez said the campus supercomputers provide him with a way to show his ideas are working before asking for time on supercomputers elsewhere.
"You have to prove how your tool works," he said.
Mahmoud Moradi is a new assistant professor in UA's Department of Chemistry and Biochemistry.
He's working to better understand human proteins through computer simulations. Moradi said he's studying a particular protein, P-glycoprotein, that's resistant to multiple drugs -- thus hindering cancer treatments.
Important protein functions take place "on a millisecond time scale," Moradi said. But a basic computer only allows for simulations of a shorter time period, Moradi said.
"When we have larger supercomputers, we can simulate larger biomolecules and for a larger time. We can basically simulate slower processes which are more relevant for biomedical research," Moradi said.
With drug development costs continuing to increase, Moradi emphasized the importance of gaining a better understanding of certain biological processes.
"The way that you look at the problem is, instead of just trial and error, we want to understand first what this protein does before we design drugs," Moradi said. But even now, the fastest supercomputers lack the ability to simulate processes over long-enough time periods, Moradi said.
As someone near the beginning of his career -- he previously was a post-doctoral researcher at the University of Illinois at Urbana-Champaign -- he said it will take time for him to write proposals to gain access to the fastest supercomputers.
Moradi said UA's computing resources helped him decide to join the faculty.
"I need good computers to do this work, and you don't have them everywhere," Moradi said.
Metro on 07/05/2015