Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.
May 9, 2008
4 Min Read
Purdue University's IT department dedicated May 5 to assembling Dell PowerEdge servers into a campus supercomputer to replace its existing research unit. It had the new machine assembled from component parts by noon, and at 1 p.m., 500 nodes were churning out results.
"There were no problems installing the hardware or software. There is no cloud to accompany this silver lining," said Gerry McCartney, university CIO, who had the idea of a staff-assembled supercomputer. "The assembly was much faster than we expected. By noon we were doing science," he said in the Purdue News Service summary of the event.
"We didn't fly in engineers or hire specialized technicians. We were able to do it with our own IT staff in about four hours," he said.
The new machine is estimated to be capable of 60 teraflops per second, or 60 trillion operations, which places the Purdue machine as number 40 on today's Top 500 computer list. The National Science Foundation estimated that supercomputer, called Steele, is capable of executing 40 teraflops on its supercomputer user's portal, the TeraGrid User Portal at http//portal.teragrid.org.
McCartney's reference to a silver lining meant, among other things, that the Purdue staff had good weather for its computer-building event. Wind and rain might have played havoc with exposed computer parts sitting outside, as 812 server boxes were unpacked under a large picnic awning in the parking lot next to the Purdue Mathematics Building. The servers were unpacked, then moved on carts to the Math building's basement.
The unpacking produced 6,000 pounds of cardboard boxes and packing and 600 pounds of Styrofoam, all of which were carried by hand to a nearby garbage truck for compacting. The university's Information Technology department had arranged for the material to be recycled, said Steve Tally, a writer for the Purdue News Service.
The process of building the supercomputer was speeded up by the help of a team of technicians sent by Indiana University, one of Purdue's arch rivals on the playing field. "We don't routinely get the opportunity to work together in person. Our meeting today was enjoyable," and may lead to future collaborative efforts, said Matt Link, director of research technology systems at Indiana.
Twenty-five faculty members pooled research grants to come up with 75% of the funding for the supercomputer, dubbed Steele for a former faculty member and director of the university computing center, John Steele.
McCartney inherited a research computer, Lear, as he came into the CIO's job. It had been preceded by MacBeth and Hamlet. He decided to change the naming convention from Shakespearean tragedies to honor former members of the department, Tally said.
One of the researchers who will use Steele looked on as the supercomputer was built. Ashlie Martini, assistant professor of mechanical engineering, will use Steele to study friction at the molecular level. Rudolf Eigenmann, professor of electrical and computer engineering, said Steele will also be used in the search for new drugs and materials, the study of weather patterns and global warming, and the design of new aircraft.
Dwight McKay, Purdue’s director of systems engineering, said Steele is running Red Hat Enterprise Linux 4.0. It’s running a number of scientific programming libraries, typical scientific applications, and the cfengine, an open source management tool for running a large cluster.
Each of Steele's nodes is a 1u server or takes up one rack mount slot, and supplies two Xeon processors, each using four-cores. That gives Steele a total of 6,496 cores, compared to predecessor Lear's 1,024 cores.
The 64-bit Xeon processors in Steele run at Intel's high end of 2.33 GHz. Each of the 812 servers is capable of having between 16 and 32 Gbytes of memory, according to the NSF site.
It is the largest supercomputer on a Big Ten campus, except for those built by the National Science Foundation as part of a National Supercomputer Center. Next door to Indiana's Purdue is the National Center For Supercomputing Applications at the University of Illinois at Champaign/Urbana.
About the Author(s)
Editor at Large, Cloud
Charles Babcock is an editor-at-large for InformationWeek and author of Management Strategies for the Cloud Revolution, a McGraw-Hill book. He is the former editor-in-chief of Digital News, former software editor of Computerworld and former technology editor of Interactive Week. He is a graduate of Syracuse University where he obtained a bachelor's degree in journalism. He joined the publication in 2003.
You May Also Like