Communications of the ACM
pgrade Triples Computational Capability of Razor Supercomputer
The expansion gives the Razor II system a total of 3,320 computer cores and a processing capability of 55.35 trillion floating point operations per second.
Credit: University of Arkansas
Student and faculty researchers at the University of Arkansas will benefit from a substantial hardware upgrade to Razor, the flagship supercomputer of the Arkansas High Performance Computing Center. The upgrade is funded by a $900,000 grant from the U.S. National Science Foundation and significantly increases the computer's computational capability, says Rick McMullen, who was recently named the center's new director.
"The Razor cluster is a major asset for U of A researchers," McMullen says. "This upgrade brings world-class computing power to the university and the state and makes us more competitive as a research intensive institution."
The upgrade occurred as a result of a multidisciplinary research project to provide better tools for computational science across the University of Arkansas campus. Four researchers — principal investigator Douglas Spearot, associate professor of mechanical engineering, and co-principal investigators Jackson Cothren, associate professor of geosciences; Magda El-Shenawee, professor of electrical engineering; and Peter Pulay, professor of chemistry — submitted the NSF grant proposal.
The original Razor system, which was activated in early 2011, featured 126 dual hex-core processor units with a processing capability of 17.75 teraflops, or one trillion floating point operations per second. The phase II expansion — now called Razor II — added 113 dual eight-core processor units with a capability of 37.67 teraflops. As a whole, the system now boasts a total of 3,320 computer cores and a processing capability of 55.35 trillion floating point operations per second.
"In other words, we've tripled our computational capability," says Jeff Pummill, manager of cyberinfrastructure enablement for the center. "To give you a sense of how powerful this is, it would take more than 40,000 years for a human to calculate a trillion floating point operations on a manual calculator."
The Arkansas High Performance Computing Center supports research in computer science, integrated nanoscience, computational chemistry, computational biomagnetics, materials science, and spatial science. Faculty and students from a growing number of departments at the university use high performance computing resources to explore the fundamental properties of chemicals and nanomaterials, complex problems in phylogenetics and genomics, new methods of detecting breast cancer, the organization of large sets of spatial data, and many other specialized problems.
For example, the computational power of Razor II helped physics professor Salvador Barraza-Lopez on a project designed to understand how pinching a graphene membrane changes its electronic properties. To do this, Barraza-Lopez and his research collaborator wanted to increase the number of atoms in a molecular dynamics simulation by more than one order of magnitude, a task that greatly increased the computational demand for the problem.
"This was possible only because I had direct access to 1,600 processors," Barraza-Lopez says. "The textures needed were generated by Razor II in less than one day, a task that would have taken about ten days of continuous computing at a normal, shared-resources facility."
Although scientific computing is indispensible in disciplines such as chemistry, physics, and engineering, researchers in other disciplines are also turning to high performance computing to meet their analysis and simulation needs, McMullen says.
"We are seeing more non-traditional users from fields such as geography, biological and agricultural engineering, and business take advantage of computing resources and expertise offered by the center," McMullen says. "We expect to see greater use by biologists, sociologists, and medical researchers as well."
The Arkansas High Performance Computing Center is funded by grants from the National Science Foundation and the Arkansas Science and Technology Authority. The center's resources may be used by any faculty member or student at the University of Arkansas or other Arkansas universities that are members of the Arkansas Research and Educational Optical Network, or ARE-ON.
McMullen, the former director of research computing at the University of Kansas, has served as director of the Arkansas High Performance Computing Center since Aug. 13. In addition, he is a faculty member in the College of Engineering's department of computer science and computer engineering.
No entries found