Lincoln (Neb.) - Jan. 17, 2002 - PrairieFire, a 256-processor computer installed this week in a new University of Nebraska Computer Science and Engineering facility in Lincoln's Miller and Paine Building, 121 S. 13th St., is estimated to be the world's 100th-most powerful supercomputer.
This machine is estimated to perform computations at a peak rate of 250 gigaflops (250 billion floating point or arithmetic operations per second), and will contain more than 100 gigabytes of collective RAM and two terabytes (2 trillion bytes) of collection hard drive storage (enough to store every book in the Library of Congress) and is 400 times faster than a Pentium III desktop PC.
The system will be dedicated to collaborative scientific computation, with priority going to research that would otherwise have been impossible to conduct using other Nebraska resources.
The Research Computing Facility, a campuswide entity that supports researchers in developing scalable, high performance code, will administer this machine. This cluster will augment and complement an existing 32-processor system. The latter machine is a more traditional architecture, consisting of a single integrated operating system, while PrairieFire is a cluster, meaning many otherwise independent machines are networked closely together via a high-speed interconnect.
"Conceptually, it is similar to the difference between a console and a component stereo system," said David Swanson, coordinator at the Research Computing Facility. "The former is easier to install and use, the latter is more powerful."
UNL faculty, computer vendors and students this week completed the installation of the system, which consists of six refrigerator-sized racks of machinery, in the downtown building. Its final cost nears $500,000, and its size is near number 100 in the "Top 500 Supercomputer sites," an international list maintained by the University of Mannheim and University of Tennessee (see http://www.top500.org on the World Wide Web).
The challenge in using PrairieFire is keeping each processor busy doing useful work and minimizing the cost of transferring information among the computers that comprise the system, said Rich Sincovec, chair of computer science and engineering at UNL.
"Computer scientists are actively involved in developing algorithms and tools to enable others to solve their problems more easily on PrairieFire," he said. "The goal is to make the hardware details of the underlying architecture transparent to the scientific programmer, much as an Internet browser simplifies searching for information on the World Wide Web. It is a small conceptual step to extend the idea of a computational cluster beyond the walls of a localized machine room to utilizing the abundant computers available via the Internet."
PrairieFire will promote inroads in this emerging area of research, known as grid computing, Sincovec said. It will enable advanced simulation to perform product analysis, design, development, testing, and manufacturing in a virtual environment. It will also provide the computational capabilities required to make advances in bioinformatics, crisis management, drought risk management, groundwater movement and remediation, combustion, and climate prediction.
A variety of work has already begun utilizing cluster computing at UNL, including molecular dynamics computations of RNA folding, electronic structure calculations of novel biological and energetic materials, and machine learning algorithms.
The use of PrairieFire by graduate and undergraduate students will be promoted in coursework, especially in class and research projects involving large-scale simulations and computations as well as in classwork involving the installation and administration of these machines.
"As a result, it is anticipated the technology behind PrairieFire will spread rapidly from downtown Lincoln, to the rest of the state of Nebraska, and beyond," Sincovec said.
The majority of this machine was purchased using funds from a National
Science Foundation grant jointly submitted by Information Services and
computer science and engineering through the Nebraska EPSCoR office.
Substantial funding has also been received from Information Services, the
University of Nebraska Foundation, Center for Communication and
Information Science, Bioinformatics Research Laboratory, department of
computer science and engineering, Center for Materials Research and
Analysis, the Office of the Vice Chancellor for Research, College of Arts
and Sciences, and the College of Engineering and Technology. A
smaller-scale prototype version of PrairieFire is located in a former
mass spectrometry laboratory in Hamilton Hall.
For questions regarding these releases, contact:
tsimons1@unl.edu
(402) 472-8514, Fax: (402) 472-7825