|
Strong connections make Pat Goda's supercomputer—and his life—greater than the sum of its parts.
Pat Goda ’93 is finishing off a green chili cheeseburger at the Hill Diner on Trinity Drive, a mile or so across the canyon from his laboratory at the Los Alamos National Labs. We’re talking about the inundation of Florida.
“What you’re seeing here is the impact from an asteroid that leaves a transient crater 300 kilometers in diameter—roughly the size of the K-T impactor, the one that caused the extinction of the dinosaurs,” Goda says as a pea-sized indentation forms on his Sony laptop computer’s screen image of the mid-Atlantic. The circle widens and spreads from the impact site and changes color, showing the height of the wave—about 1,000 feet high 100 miles from the impact site and moving at 400 miles per hour.
Our waiter leans over the table to observe the impending disaster.
After the initial strike, the crest of the wave fades and the tsunami becomes a huge, rolling swell. If you’re at sea 1,000 miles from impact, you can ride it out. But about three-and-one-half hours later, Bermuda is gone. Little more than five hours after impact, the ocean at beaches on the U.S. east coast recedes hundreds of yards, then runs up in a huge tidal surge. There’s no monster wave, but a relentless rising of water (at Miami, to 200 feet above ground level) with storm-sized, debris-laden waves. Six hours after the asteroid hits, the cities of the east coast and most of Florida are underwater.
Bruce Hornsby’s “The Way it Is” plays as Goda pulls out a color-coded map showing the various heights of the wave during the entire event. Astrophysicist Jack Hills, who led Goda’s research team on asteroid impact and was the Wabash physics major’s mentor when he arrived at Los Alamos as an intern in 1992, calls this one “the finger of God.”
The waves Goda models are hypothetical. Asteroid impacts the size of the K-T impactor occur once in 65 million years, and smaller strikes the size of the asteroid that leveled 800 square miles of forest near Tunguska, Siberia in 1908 hit once every 300 years. But Hills’ and Goda’s simulations have real-world implications, whether the source of the tsunami is an asteroid impact, an underwater landslide, or an earthquake.
Goda can’t help thinking about those implications and wants to model them, as well.
“What we really want to know, of course, is the potential for loss of life,” he says. “There are all sorts of sociological and psychological questions: How are people going to react? How fast are they going to be able to evacuate when they find there’s been an impact like this?
“Then there are policy questions: What do we want to do? Do you invest money into first finding these things and destroying them before they hit? Are we going to mitigate the hazard, or are we going to react when it happens? It’s a matter or priorities.”
Our waiter takes away our plates and brings the check.
“Did you know that New Mexico actually has a State Question?” Goda asks. “It’s ‘Red or Green?’”
“This is a bad state to live in if you don’t like your food spicy.”
The simulations on Goda’s laptop aren’t Hollywood-hype Deep Impact speculation (although Jack Hills was consulted by that film’s producer). Goda’s S.W.I.M. (Shallow Water Immersion Model) is about as close to a “real world” simulation as scientists can get without dropping a hot rock into the ocean. Goda and his colleagues wrote all the factors and variables of an asteroid impact—the mineral make-up of the asteroid and its speed and trajectory on impact, the topography of the ocean floor, the contour of the continental shelf, the temperature of the water, and all the laws of physics and of fluid dynamics—into the program that drives the simulation.
Running such complex models demands the sort of supercomputing power that’s hard to come by at Los Alamos when you’re an astrophysicist in the non-weapons-related Theoretical Astrophysics Division, or T-6.
So Goda—who took only one computer science course at Wabash—and T-6 colleague Michael Warren, an astrophysicist whose work includes simulating the origins of the universe, made their own supercomputer.
Their creation is having a deeper impact on the world than any asteroids are likely to have in Goda’s lifetime.
In 1996, Warren and Goda built Loki and won the prestigious Gordon Bell Prize for practical parallel-processing research. Costing about $50,000 and based on the “Beowulf” parallel-processing model developed by do-it-yourself supercomputer pioneer Thomas Sterling, Loki clustered readily available hardware—16 Intel Pentium Pro processors—connected with Fast ethernet technology and a Linux operating system in a way that created a computer exponentially more powerful as a scientific tool than a single supercomputer. The whole was greater than the sum of its parts.
Warren took the feat up a notch in 1998 with the Avalon cluster, the 315th fastest computer in the world and built with a price tag of $150,000—a fraction of the cost of most supercomputers.
“Researchers at Los Alamos have startled the supercomputing world by building in-house one of the fastest computers ever at such a modest price,” Wired news reported. “By making supercomputer power and assembly cheap and accessible, control over the world’s very fastest machines shifts from a few supercomputer vendors to a much broader community.”
It was quite an achievement for astrophysicists who were really just looking for a way to run their simulations. But as their models have become more complex, their quest for power has continued. In November 2002, Warren, Goda, and Chris Fryer brought online their latest cluster. This time they used generic versions of the game box your kids play with to create one of the world’s 100 fastest super- computers. They call it “The Space Simulator.”
An old Gordon Lightfoot song is playing at the Hill Diner as Pat Goda tries to explain how a Crawfordsville kid with an inordinate interest in the weather came to build one of the world’s fastest supercomputers.
“Science was my whole life when I was a kid,” he recalls. “In third or fourth grade I read that you could take a battery charger and aluminum foil and split water into hydrogen and oxygen.”
So Goda filled a milk jug with hydrogen, made a detonator fuse out of the foil, and hooked that up to the battery charger.
“I blew the thing up.” Goda laughs. “My mom thought this was not quite normal.”
He studied weather at school and in 4H, writing frequently to Indianapolis weatherman Bob Gregory, who sent him forecast maps. In high school he made a conscious decision to attend a small, liberal arts college.
“I decided that the liberal arts was a more rational approach to life than what the big schools had to offer,” Goda explains.
He gained from Wabash an intellectual agility that continues to serve him.
“At Wabash, I learned to learn,” Goda says. “And that’s what I teach when I have students here at the lab. I don’t teach them to memorize or specialize, but how to go out and find the answers themselves, and how to question the answers they find.”
“And that’s the nature of research,” Goda says. “We’re not just finding answers; we’re looking for something original, and we have to be a little creative.”
A little “creativity” won Goda an internship that gave him his first contact with Los Alamos and Jack Hills.
“It was down to the final cut and they asked me if I knew FORTRAN,” Goda recalls with an embarrassed blush. “I knew Pascal, I was good at picking up computer languages, so I said, ’yes,’ trusting that I could learn it quickly.”
Hills was surprised when his student intern showed up at Los Alamos still studying the language. But Goda delivered, published with Hills a significant paper on asteroid impacts, and when Goda finished his graduate work in meteorology and was looking for work, Hills was quick to hire him at T-6.
“I’m in theoretical astrophysics now, but the thing that ties together all of the science I’ve done is fluid dynamics,” Goda says. “That’s how meteorology, meteoretics, and astrophysics are all tied together. And nine times out of ten you can’t do an interesting problem in fluid dynamics without a computer.”
“We’re going to try to do the biggest 3-D simulations ever,” Goda says as he leads me toward Room 157, where The Space Simulator cluster supercomputer resides. The 3-D simulations of supernovas began running in November 2002. Godas’ asteroid impact simulations will come later.
“The system should come in somewhere between 70 and 80 in the top 100 fastest,” he adds, but Mike Warren shakes his head.
“More like 110, 120,” he says.
The carpet in the hallway is worn and stained. We step into the cooled room where the computers live and see boxes holding cables and keyboards. Other parts lie scattered about. It looks like a high-tech Christmas morning in a teenager’s room, where the excitement of seeing how it all works far outstrips any concern for tidying up. In T-6, image is nothing—the work is everything.
The Space Simulator is 294 separate game boxes, each costing less than $1,000 and stacked 12-high on metal racks that Goda purchased from a food service company. The cubes are connected by one-inch diameter cables that look like they’re filled with blue ice, and the temperature as you walk between the stacks is between 69 and 75 degrees, monitored 24/7. Goda can give you the heating/cooling calculations from memory, likewise the cost of each box, the type of ethernet card used; he can tell you where he scrounged the antique Dell monitor and the yellowed, 20-year-old Fujitsu keyboard that commands this world-class supercomputer. Both set on an old typewriter stand.
“Hey, it works.” Goda smiles. For a man whose motto has been, “I may not know the answer, but I know how to find it,” building The Space Simulator has been a most satisfying challenge.
“I was constantly learning,” Goda says as we step out the doorway, the only sound behind us is the hum of the eight-foot-high air conditioning unit. This supercomputer that will be pondering the origin of the universe does its thinking quietly, and in a cold room.
“She’s a nuclear engineer. I’m in theoretical astrophysics. It’s a good arrangement. I know in theory how to fix things in the house, and she knows how to do it.”
Goda’s talking of his wife, Joetta, whom he met in 1996 and married in 1998. The townhouse they share in Los Alamos has the lived-in feel of a couple with diverse but compatible interests. A jazz saxophonist at Wabash, Pat still plays music with friends. Inspired by B.B. King and Wes Montgomery, he practices jazz guitar in the upstairs loft. (“Jazz improvisation is just like science—you learn some rules, then you see how far you can stretch them.”) Joetta likes to sew while he plays. A garden out back provides a place to work together, the family cat a creature to dote on and monitor with a ceiling mounted “cat-cam.”
“The most important thing to me is my wife, my family, and then my own well-being, and somewhere down there are job and my career,” Goda says. “Taking that as a place to start, I feel a lot more freedom in my career.” He feels no compulsion to earn his doctorate. He enjoys his work at Los Alamos, and though he’s not sure what kind of science he’ll be doing in the long run, he senses science and computers will always be part of his life.
“I was born in the same year that the first single-chip processor—the Intel 4004—was invented,” he says, admitting that he has a processor- chip collection.
“I know—the ultimate geek hobby,” Goda says. He can’t show me the set—his in-laws are mounting the collection in a shadow box for him as a gift. He speaks of them fondly, as he does his friends, colleagues like Hills and Warren, even the diner we just left.
“I love that place,” he says, and when I tell him it reminds me of a large-scale Scarlet Inn, he says he spent a lot of time there as a student.
“I knew the people there, and I used to take my books there, especially for C&T, and read— that’s where I did my homework. And you could go down there and pick up bits of conversation between the professors.
“My wife and I come here a lot now, and the lady who works here in the evening knows us, and she always gives us a booth; she knows we like to sit over here. It’s really nice to have a place like this in town.”
Warren and Goda were both wrong: The Space Simulator placed 85th in the top 500—the 85th fastest computer in the world is made of game boxes!
His computer, drawing from so many sources, so quick on its feet, built by a liberal arts grad, and with a power that’s greater than the sum of its collective parts, is a decent metaphor for liberal arts education. Or Goda himself.
He’s determined not to be driven by any one thing, but to balance his interests and his passions, remain mindful of all he’s learned and experienced, and to reflect on the present and all its possibilities. When you’re a scientist in the town they created to build the atomic bomb—where the main street through town is Oppenheimer Drive and you can’t pass by the town’s “Trinity on the Hill” Episcopal Church without a pinch of irony—you’re wise to keep your mind awake.
Goda does. I ask him about the forest fires that threatened the town and the labs last year.
“The smell was in the air for months, but emblazoned on my memory from that time is the color that shadows cast when the fires were near—not black, but blue.”
Of course, the scientist had to know why.
“It’s called Mei scattering,” he says. As he explains the theory, I realize that for the first time in a conversation that’s spanned several hours and engaged politics, ethics, computer science, geology, and economics, among other subjects, we’re actually talking pure physics.
|
