Science Gets Boost From Supercomputers : Technology: Supercomputers are substituting as laboratories to study complex or inaccessible phenomena of nature.
Inside the world’s fastest computers, rows of sweet summer corn are being attacked by insects. Thunderstorms are brewing. Hearts are beating. And stars are colliding.
Although they undulate with energy and follow the laws of nature, none of these activities are real. The insects and thunderstorms are nothing more than numbers and equations rushing through the processors of a Cray-2 or ETA10 Model E supercomputer.
Yet to a growing number of academic and industrial scientists, the colliding stars and beating hearts are real enough. More and more, supercomputers are being used as laboratories in which to simulate and study complex or inaccessible phenomena, such as the churning innards of a storm or the onset of a heart attack.
“The supercomputer allows us to do experiments that would be extremely cumbersome, if not impossible, to do in the lab,” said Art Winfree, a biologist at the University of Arizona who is trying to understand the process of cardiac arrest by doing computer experiments at the John von Neumann National Supercomputing Center at Princeton, one of five such facilities supported by the National Science Foundation.
Like most scientists using supercomputers, Winfree has not abandoned studying real life. But he has found that by using a supercomputer, he can examine in minute detail the swirling bioelectrical vortex thought to be responsible for the onset of ventricular fibrillation in heart muscle, the irregular contractions that precede a heart attack.
Doing the same studies in a human heart is not yet possible. Doing the same simulations on Apple or IBM personal computers, instead of a supercomputer, would take centuries or longer.
The supercomputers are so big and so fast that for the first time researchers can simulate such feats as the head-on collision of superdense neutron stars and the intertwining of cosmic strings. They can simulate black holes and drop matter into them. They can watch model galaxies spew jets of gas into interstellar space.
On a more practical level, supercomputers can be used to make a better ice-cube tray by enabling designers to understand how hot plastic fills a mold. Engineers can design better bridges by investigating how cracks propagate through different materials.
David Onstad of the Illinois Natural History Survey has studied the population dynamics of insects. At the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, Onstad recently simulated the intricate life histories of a tiny beetle that infests grain silos. On the supercomputer’s circuits, as in the real world, the beetle was attacked by microorganisms and subjected to cannibalism. By doing the millions of computations necessary, the supercomputer allowed Onstad to understand how beetle populations over hundreds of days are controlled by disease and predators.
The supercomputer’s speed and memory are the sources of its allure. Calculations that would take months or years on the ubiquitous work station can be done in minutes or hours on a supercomputer. Onstad, for example, recently simulated how a population of European corn borers changed over 25 years. The program took only hours to run through the supercomputer, though reviewing the data and presenting it took several more months.
The machines achieve their speed by doing many calculations simultaneously, a trick called “vector processing,” in which several rows or “vectors” of equations are attacked by the computer at the same time. Even faster computers are now being produced that do “parallel processing.” Compared to these new machines, ordinary computers are electronic dullards--doing their work step-by-step, completing one task before moving on to the next.
This ability to do so many calculations gives Robert Wilhelmson, a meteorologist at the University of Illinois, the ability to simulate severe thunderstorms within a computer. Into the supercomputer, Wilhelmson feeds the basic equations that describe how fluids are transported and clouds are formed. He then gives the computer some other basic parameters, such as temperature and moisture. Then he simply turns the program on.
“We can use the computer to study the dynamics of the storm, and we can see things that we can’t observe in the real world because things are happening too fast or they’re too hard to sort out,” said Wilhelmson.
Indeed, Wilhelmson confesses that he is fascinated by the images that the supercomputer and its advanced software can create. While most scientists are eager to show visitors their slides and graphs, Wilhelmson preserves his experiments on videotape.
On his tape, a small cloud gathers strength and grows in size, forming broad shoulders and then becoming dark and threateningly anvil-shaped. As the video runs, and while Wilhelmson stops and starts and rewinds the tape, the simulated thunderstorm shows how updrafts and downdrafts influence the storm and how wind circulates inside the cloud.
Wilhelmson believes that being able to visualize the computer simulation is crucial to understanding the experiments. “You are producing gigabits of data,” he said, referring to the supercomputer’s capacity to perform functions involving billions of individual digits. “So much data is coming out of the computer that we have to be able to move it around and make sense of it. You have to find some order in all this,” he said.
To deal with the volume of data, scientists have turned to artists and software engineers to produce three-dimensional graphics of the research. Much of this software was first developed by technicians working in the entertainment and advertising industries. Indeed, some of the most advanced software programs come from the people who produced movies loaded with special effects.
“These visual images allow us to extract scientific knowledge that is hidden in all these data,” said Karl-Heinz Winkler, deputy director of the National Center for Supercomputing Applications at the University of Illinois. Winkler said this visual information, not numbers that a computer can understand, are at the heart of understanding phenomena such as thunderstorms and stellar collisions.
Still, some researchers worry that the pretty pictures generated by computers do not always tell the whole story. “A lot of the complex computer graphics you see are closer to art than to science,” said Donna Cox, an assistant professor and artist at the University of Illinois who works with scientists to create images of their computer-generated experiments that are both accurate and insightful.
“There’s more computers around, and more scientists are bowing and scraping before them and believing what comes out of them as if they were some kind of oracle,” said Winfree. “You have to remember that 90 percent of what comes out of a supercomputer is junk. You have to remember it’s a tricky business. You have to be careful.”
