Natural Product Chemistry, Applied Mathematics and Aerodynamics Research Get Performance Leaps from Shared-Memory Architecture

MOUNTAIN VIEW, Calif., Aug. 22 /PRNewswire-FirstCall/ -- To safely develop new drugs from organic compounds and to develop more stable aircraft and satellites by harnessing the compute power of shared-memory architecture from Silicon Graphics (NYSE: SGI), the Office of Information Technology and the Office of Research as well as the College of Science and the Mathematics Department at Virginia Tech, in Blacksburg, VA, recently selected SGI(R) compute and visualization technology. The SGI systems will be used as centralized resources available to all faculty and researchers, and to easily port a host of scientific codes to the open systems Linux(R) OS.

A new SGI(R) Altix(R) high-performance computing system, purchased in April, is up and running and almost 100 percent booked by faculty researchers and graduate students, primarily in applied mathematics, high-altitude aerospace and mechanical engineering, and quantum chemistry, which is aiding drug discovery. The University has already placed limits on the users to enforce fairness and plans to implement a queueing system in the future that would keep the machine at 100 percent utilization while allowing researchers access to as many CPUs as they need for their program. The Silicon Graphics Prism(TM) visualization system will be configured as part of an immersive visualization theater environment for a variety of scientific research.

One of the most important things the University's Research Division required was the ability to port a variety of codes developed by faculty members, in addition to community-based codes developed through a consortium of researchers around the world, as well as publicly available domain codes.

"What shocked us is the ease with which the researchers were able to get applications to execute in the Altix system's large, shared-memory environment," said Erv Blythe, vice president of information technology, Virginia Tech. "Researchers value their time probably as much as anything, and in terms of setting up their code, they care about good utility, getting good returns for that investment in time, and reliable and stable computing. Researchers who are used to working on clusters are seeing a stark difference in terms of reliability and stability with Altix. From my perspective, this looks like one of the best investments we've made in research computing in the last several years."

irginia Tech, which consistently ranks among the top research institutions in the U.S., had purchased a 16-processor SGI Altix system a year ago. Researchers loved it; almost immediately, the system was over-booked by 400 percent or more. In April 2005, Virginia Tech, purchased a SGI Altix high-performance system with 64 Intel(R) Itanium(R) 2 processors and 256GB of memory through SGI's Higher Education and Research VAR James River Technical, Inc. The University also purchased, in the same month, a Silicon Graphics Prism(TM) visualization system with four Intel Itanium 2 processors, 4GB of memory and two ATI(R) graphics processors, to jump-start a scientific visualization initiative. Both systems run the true 64-bit Linux environment.

"There are two main advantages of SGI Altix," said Kevin Shinpaugh, director of research and cluster computing, Virginia Tech. The first is the high-performance 64-bit processors. We went from a system that was over five years old to the Altix. Essentially, one of the Itanium 2 processors that we are using now is equivalent to all 16 of the processors in the older system, in terms of performance. The second advantage Altix offers is the wide variety of scientific and engineering programs that is available-or can be made available-for the system because of Linux. There are so many different programs that different departments need, and the Altix was really the only system that had the capability to run these programs and run them very fast."

Quantum Chemistry To Develop New Chiral Drugs

In the drug research area known as natural product chemistry, samples of organic compounds from around the world are collected and scientists are frequently able to develop new drugs based on these compounds. One such drug is the anti-cancer agent paclitaxel (patented as Taxol), which is actually derived from the bark of a Pacific yew tree. More specifically, it is derived from one "hand" of a chiral molecule taken from yew tree bark, and then synthesized in a laboratory. It is critical to precisely match the "hands." Using the wrong hand of a chiral molecule, or using the entire molecule, can be tragic. It was discovered that infant deformities resulting from Thalidomide use in the late 1950s/early 1960s came from distributing the drug's entire chiral collection. Today, most drugs we take are a single chiral orientation.

Dr. T. Daniel Crawford, associate professor, chemistry, at Virginia Tech, was one of the first professors to leap at the chance to study chiral molecules on the new SGI Altix system. "In quantum chemistry we carry out fairly large scale modeling of individual molecules using the principles of quantum mechanics," said Dr. Crawford. "These kinds of calculations tend to be pretty intense, and they really stress almost all parts of the computer. We require systems to be very fast, with very large system memory, and very large disk with fast I/O. The SGI Altix system has all of these features. The very large memory is especially appealing because the complicating factor in chiral molecules is that you may actually have not just two possible hands, but a thousand possible hands. The Altix will allow us to do much larger calculations than we've ever been able to do, so far."

Applied Mathematics for Aircraft and Space Systems

The Interdisciplinary Center for Applied Mathematics at Virginia Tech brings faculty from various disciplines together to work on research projects. Terry L. Herdman, director of the Center, and also director of research computing, works with a group that is using partial differential equations to build accurate mathematical models that are firmly grounded in physics and engineering principles, which, in turn, lead to analysis and simulations of aircraft and space systems. Most noted for its study of flutter of aircraft wings, researchers developed rigorous mathematical models for the unsteady aerodynamic loads and proposed feedback laws to control flutter.

"Our current project, which is a joint effort with DARPA, NASA Langley and the Air Force Research Labs, is to develop accurate mathematical models for large space systems," said Herdman. "In particular, we're looking at building a large space antenna, being able to deploy it in either low- or mid-Earth orbit and then being able to control it, point it, etc. If there are any vibrations in it, we will be able to mathematically model those parts and determine how and where we can control them. The SGI Altix is a tool to enable us to provide approximations for solutions to these space systems, and also give us simulations. Altix also makes it easy to use third-party software. Simply put, it is a system that's reliable, very easy to get on to, to run, to get successful runs and to be able to look at results in a timely manner."

"The research community in higher education is a key driver in leveraging the power of today's science and technology and the sharing of ideas to create the scientific and technological breakthroughs of tomorrow," said Dave Parry, senior vice president and general manager, Server and Platform Group, SGI. "The performance of SGI's NUMAflex architecture, coupled with the portability and ease of use inherent in its open-standards Linux operating environment, are important factors in enabling these breakthroughs. We at SGI look forward to seeing the fertile minds at Virginia Tech use their new Silicon Graphics Prism and Altix systems to produce advances that benefit society and humanity."