UK Researchers Surging Ahead With High-Speed Computer Networks
"We are made of star stuff," says Gary Ferland, a University of Kentucky professor of physics and astronomy, as he reaches across his desk for a nearby book. He opens it to a stunning photograph. A star, millions of light years away, is pushing gases into space, the result of a nuclear reaction taking place inside the star. The reaction creates a billowing cloud of yellow-gold blossoming against the black backdrop of space.
While most of us merely marvel at the beauty of such an image or the ability of science to photograph it, astronomers such as Ferland use these images to understand complex processes taking place in the galaxy. By studying the light, scientists can determine the age of stars. They can even tell us that our skin contains some of the very atoms found in such stars. Ultimately, they hope to answer questions that have forever intrigued mankind -- questions no less weighty than where we came from and how the universe was created.
Unlike other scientists, however, astronomers cannot conduct hands-on experiments to further their research -- nobody can do an experiment with a star billions of light years away. Astronomy is an observational science. And now, in addition to instruments like the Hubble Space Telescope, Ferland and his fellow stargazers can simulate conditions in the galaxies with computers and then compare those simulations with other data from the heavens.
These simulations require massive computing power and an ability to communicate via computer with other researchers around the world. It is precisely the sort of computing challenge that excites John Connolly, director of UK's Center for Computational Sciences.
Connolly came to UK when the center was founded in 1987. Governor Martha Layne Collins established computational sciences as one of only five Centers of Excellence in the state, and a special $5 million appropriation from the Kentucky legislature that same year provided funds for the University's first supercomputer.
In the 10 years since that time, the center's mission has been to support researchers in multiple disciplines across campus by meeting their needs for a computer network able to process computations several hundred times too complicated for even the most high-powered personal computer. The center also funds fellowships for professors and graduate students doing research that requires cutting-edge computing.
Expanding the Horizons of Knowledge
According to Connolly, the biggest demand for the center's computational services typically comes from Ferland's counterparts in hard sciences such as physics or engineering.
"This is the kind of thing a university should be doing, expanding the horizons of knowledge," Connolly says of Ferland's work. "It is a good example of pure science, interesting in and of itself. It extends our knowledge and keeps us in close contact with what others in the field are doing. It keeps us up to date."
John Connolly, director of the computational sciences center, says that with the new UK library, the large computers in the center and the expanded networks, "It's all coming together."
Still, these scientists are far from the only beneficiaries.
Tucked away in a corner office on the 13th floor of Patterson Office Tower, two computers hum on the desk of another scholar, Kevin Kiernan, a professor in the Department of English. Since 1993, Kiernan has been immersed in a project that will soon result in a two-volume compact disk called The Electronic Beowulf.
As he opens a test version of the disk, a page of the ancient poem soon appears on the computer screen. The page is a digitized copy from the only original manuscript of Beowulf. The Old English letters are there to examine. The color and texture of the manuscript's vellum leaf is visible. Even the tape used to attach the pages to a paper frame can be seen.
Perhaps most significant, the edges of the manuscript are visible, revealing hundreds of letters or part of letters covered by frames which were designed to preserve the fragile text after the edges were seared in a 1731 fire. Even the few researchers previously allowed access to the manuscript could view the hidden letters only if they held the page up to bright light.
Kiernan's work -- made possible by the center's computational support -- will change virtually everything about the way scholars study what is considered the most important work of Old English literature. Not only will they be able to read the original 11th-century manuscript electronically, they will be able to examine copies made by two scribes in the 18th century, as well as 19th-century collations and editions.
"Now, for example, scholars interested in the construction of the original gatherings of the manuscript will be able to place once conjugate leaves side by side again, or examine in great detail the color and texture of the vellum leaves by magnifying the images," Kiernan writes in an article on a Web site about the project. "Anyone interested in the accuracy and diligence of the scribes, moreover, can investigate all of their erasures, which have been scanned both in bright daylight and with the sometimes more penetrating aid of an ultraviolet lamp. And, with the help of image processing programs, students will even be able to restore or at least improve the legibility of faded passages. Readers of the electronic facsimile will thus acquire a reproduction of the manuscript that reveals more than the manuscript itself does under ordinary circumstances."
A Marriage of Scholarship and Technology
Neither scholarship nor technology alone could have achieved these results, a fact not lost on Kiernan nor Connolly.
"The Center for Computational Sciences helped us with research support," Kiernan says. "They also showed us ways to deliver all these images. Their role was crucial. Without their help, none of this could have happened."
The project is a good example of the difference between computational science and computer science, a distinction which eludes many and is a perennial question for Connolly and his team.
Kevin Kiernan, a professor in the Department of English, has been immersed in a project made possible by Center of Computational Sciences support that has resulted in a two-volume compact disc called The Electronic Beowulf, a digitized-copy from the only original manuscript of the ancient poem.
"Computational science is any kind of science that uses a computer," Connolly says. "Computational science uses the computer as a tool. The computer is just a device for getting your work done, doing your calculations. Computer scientists are actually interested in the machine itself, in designing new machines and making them work more efficiently through better software, for example. They are designing the next generation of machines that computational sciences will be using five years from now."
The Beowulf project is also an excellent example of the interdisciplinary nature of the center. In fact, Connolly delights in bringing together experts from different disciplines.
Kiernan, for example, struggled with assembling the computer codes that would make it easy for people to look at different pages of the manuscript using the various tools. The center provided money for a computer science graduate student who worked with him to develop the tools for looking at manuscripts.
"The graduate student will get a dissertation out of it," Connolly notes. "What he is doing is developing the kind of tools that future libraries are going to use."
NSF has recently awarded $830,000 to the South East Partnership to Share Computational Resources (SEPSCoR), a program run by the computational sciences center. The grant will connext six campuses--Alabama, Kentucky, Louisiana, Mississippi, South Carolina and West Virginia--with a 45 megabit/second ATM network. ATMS provide transmission speeds 10 or more times faster than earlier networks.
This multidisciplinary, interdepartmental approach helps the center vault typical administrative barriers.
"What we try to do is get different levels of expertise together, kind of serve as a broker," Connolly says. Somebody from one department may have skills that someone from another department needs. In the ordinary structure of a university, you can't support that. The English department can't pay the computer science department or vice versa, but I can do that. I can pay the salary of a computer science student to work on a project in the English department. You have to have an interdisciplinary center to do that."
Multiple Disciplines, Common Computing Needs
The need for serious computing power crosses all disciplines. The huge size and complexity of computer files is a common denominator in all the research projects, whether the scholars are studying galaxies or literature. For example, Kiernan used a Roche/Kontron ProgRes 3012 digital camera to scan text at 2,000 x 3,000 pixels in 24-bit color. The resulting images are enormous, nearly 22 megabytes apiece. They require at least 64 megabytes of RAM to process in real time.
Across campus, James McDonough, an associate professor of mechanical engineering and mathematics, is doing research that is literally worlds apart from that of Kiernan's in every regard but one: McDonough's research requires comparable computing power. Applying the principles of computational fluid dynamics, McDonough and Kozo Saito, a UK professor of mechanical engineering, are seeking to improve the current method for painting cars.
By way of background, McDonough explains that all auto manufacturers lose around 50 percent of the paint they try to apply to the car shells. The cars are painted in a painting booth by robots which direct spray guns filled with paint at the shells. The paint particles which do not attach to the car are blown into a trap beneath a grated floor and then further trapped by filters so they do not escape into the atmosphere. It is an expensive process. Not only do they lose the cost of the paint, the manufacturers also have to continually replace the expensive filters.
James McDonough, an associate professor of mechanical engineering and mathematics, is applying the principles of computational fluid dynamics to improve the current method of painting cars.
Toyota Motor Manufacturing North America turned to the UK researchers to develop a better system. Now three years into the long-term project funded by Toyota, McDonough and Saito are doing computer simulations, mathematical calculations and laboratory experiments that will ultimately result in a more cost-efficient painting system. It is not unlike the research McDonough has done for General Electric Aircraft Engines in which he used computational fluid dynamics to analyze the air flow inside jet engine turbine blades, to ultimately increase the fuel efficiency of the jet.
Continually Upping the Ante
Computing power is clearly the name of the game when it comes to research dependent on large computations, and the center has been quite successful at attracting the kind of grants that allow it to continually enhance the computing power available to UK researchers.
In May of 1997, the National Science Foundation (NSF) announced that UK was one of 35 institutions allowed to connect with NSF's very high-speed Backbone Network Service (vBNS). In addition to providing connections 1,000 times faster than a personal computer, the network is dedicated to the researchers. The center supported that effort.
In September of 1997, NSF awarded $830,000 to the South East Partnership to Share Computational Resources (SEPSCoR), a program run by the center. The grant will connect six campuses -- Alabama, Kentucky, Louisiana, Mississippi, South Carolina and West Virginia -- with a 45 megabit/second ATM (asynchronous transmission mode) network. ATMs provide transmission speeds 10 or more times faster than earlier networks, and this connection will be reserved for use by researchers at campuses in these six states. The Electronic Beowulf will be a primary test vehicle for the SEPSCoR network.
For most of its existence, the center has also received grants to run a program called ESPCoR, a federal/state program designed to help rural states, typically far from major computer nodes, develop top research programs. UK was the only ESPCoR state to receive a recent grant through NSF as one of 10 major partners with the University of Illinois to contribute to a National Technology Grid. The Grid will allow computational science centers all over the country to share resources.
"This is like a huge cooperative," Connolly says. "It is brand new in science. It has never been done before."
"Never been done before" is simply a way of doing business for the center, particularly since the researchers have had access to the center's HP-Exemplar computer, one of the best supercomputers on the market, according to Connolly. The Exemplar is a parallel computer, providing both speed and flexibility.
Both are often needed. Consider the work of Robert Lodder, an associate professor in the College of Pharmacy. Lodder is studying atherosclerosis (hardening of the arteries), particularly its role as a contributing factor to strokes. His work modeling arterial plaque requires billions of computations and high-speed communications.
Using noninvasive cameras, Lodder obtains transarterial three-dimensional images of the different plaques in arteries and then does chemical composition profiles of arterial lesions at various stages of development. These data are then correlated with the patient's results from duplex ultrasound and from lipoprotein extraction and separation of the patient's plaques. It is an innovative approach, studying the lesions found in artery walls rather than the tissue removed from the site.
"Monitoring cholesterol in the blood doesn't tell you much about the progression of disease," he says when explaining why he has taken a different path than other scientists. "The action is in the artery walls, where there is a much higher concentration of cholesterol."
The research is also innovative in terms of its use of technology. Lodder is taking advantage of new imaging algorithms for massively parallel supercomputers as well as new intra-arterial fiber-optic cameras.
In his study of how arterial plaque contributes to atherosclerosis, Robert Lodder requires billions of computations and high-speed communications.
Lodder says the chemical analysis of the lesions should help researchers understand lesion formation and growth, ultimately leading to better treatment programs and a reduction in the number of people -- now more than 500,000 annually -- who are affected by stroke.
In one way or another -- whether saving lives, money or rare manuscripts -- all the research improves our lives, and that is the point of not just the research but of the center as well, according to Connolly. In 1997, the center observed its 10th anniversary, a cause for celebration, introspection and comments about the future.
"It is all coming together," Connolly says. "We are getting all of the components including the new library. We have the computational science center well established. We have the big computers downstairs. We have the networks. Those are the components of the future university -- large computational resources, good networks and good repositories."