Fayetteville, Ark. — Further demonstrating its position as a national leader in high-performance computing, the University of Arkansas has received an $803,306 equipment grant from the National Science Foundation to purchase and assemble the university?s second supercomputer. The new machine will be among the fastest supercomputers in all U.S. academic institutions.
The computer will enable new and ongoing projects in computer science, physics, chemistry and other areas. Perhaps more importantly, by connecting to networks and other high-performance computers, the supercomputer will significantly augment the state?s computing infrastructure and facilitate collaboration among researchers at other institutions within the University of Arkansas System.
?In scope and scale, this project really extends far beyond the Fayetteville campus,? said Amy Apon, professor of computer science and engineering and the university?s director of high-performance computing. ?We see its relevance as a critical piece of infrastructure to implement a statewide strategy of high-performance computing that will benefit applications in engineering, science, government, business and agriculture.?
Although many technical specifications have not yet been determined due to the possibility of harnessing imminent and significant improvements in computer-processing technology, the new machine will operate at a speed of six teraflops, approximately four times faster than Red Diamond, the university?s first supercomputer and currently the only supercomputer in Arkansas.
Assembled as a cluster or rack of integrated processors, the machine will be used for massive data storage and complex computations. In addition to supporting ongoing projects, the new computer will perform large-scale data mining, which is the process of retrieving data and identifying specific patterns in various data sets.
The new computer will connect to Red Diamond, the Arkansas Research and Education Optical Network and computer networks at the University of Arkansas for Medical Sciences and the University of Arkansas at Little Rock, which has begun the process of obtaining its own supercomputer. These connections, Apon said, will create a partnership in which researchers at all three institutions will have access to resources and be able to run applications at all sites. The new supercomputer will also be connected to the Great Plains Network and the Southern Universities Research Association?s computer grid.
As a related development, an external advisory committee consisting of a team of experts in high-performance computing has been established and will visit Arkansas in late October to help develop a strategic plan for computational infrastructure in the state. Apon said the plan and infrastructure will promote collaboration between academic and industry partners. The advisory committee is funded jointly by the University of Arkansas, UALR, the National Science Foundation and the Arkansas Science & Technology Authority.
Supercomputers solve single problems at rates measured in teraflops, sometimes called ?TFlop/s.? The term flop/s, which stands for floating-point operations per second, refers to a complex mathematical calculation to measure the speed at which a computer can process data. ?Tera? stands for one trillion, so supercomputers produce one or more trillion computing operations, or calculations, per second. There are also computing speeds of kiloflops (1,000 floating-point operations per second), megaflops (1 million) and gigaflops (1 billion).
In research, supercomputers solve ?Grand Challenge Problems,? the name given to problems that cannot be solved in a reasonable amount of time with ordinary computers. An example of a Grand Challenge Problem is global weather modeling. The U.S. government uses a supercomputer to build three-dimensional models of cubic sections of the earth?s atmosphere to analyze and forecast weather patterns. The models require processing massively large amounts of data. With a personal computer operating at one gigaflop, it would take 10 days for researchers to forecast weather over a seven-day period. With a supercomputer, researchers can calculate the same data in less than an hour.
In 2005, the University of Arkansas obtained Red Diamond, which has helped researchers solve challenging scientific and engineering problems, such as calculating the molecular structure of new nanomaterials, calculating molecular formulas for new drugs, predicting the behavior of tornados and volcanoes, and storing and retrieving massively large amounts of data. Red Diamond operates at a speed of 1.379 teraflops.