Iowa State University

Inside Iowa State
Oct. 24, 1997

From gizmos to replica

Gustafson has spent a lifetime building things

by Skip Derra

Taking a break from the full-speed-ahead world of high- performance computing, John Gustafson helped rebuild the world's oldest computer -- the Atanasoff-Berry Computer replica. But Gustafson, a computational scientist at the Ames Laboratory and an expert on high performance computing and computer benchmarking, had to rely on a lifetime of building gadgets and gizmos to help him through the project.

In the end, it was not a clever use of electronics or a detailed understanding of what John Atanasoff and Clifford Berry did 60 years ago that got the replica to work before national media in Washington, D.C. It was fine-tuning.

"Like a piano, the computer needs to be tuned," he said of the 750-pound, whirring and clanking machine.

Gustafson should know. He's built, and tuned, both.

The ABC replica is an authentic working model of the original Atanasoff-Berry Computer, built at Iowa State from 1939 to 1942. While it might appear to be far removed from the cutting edge of computer technology, where Gustafson customarily resides, it's really closer than it appears.

The ABC, Gustafson said, not only was the world's first electronic digital computer, it also was the first parallel processor. And building the replica was as exasperating as building a modern computer.

But building complex electronic instruments began early in Gustafson's life. A child who forsook baseball mitts for Heathkits, Gustafson early on exhibited an Edisonian inquisitiveness.

"I was electronically precocious," he said, "and I had indulgent parents."

Gustafson built and brought to school a radio transmitter -- when he was in the first grade. He built his first computer at the age of 13, and at a time when most boys have their heads under the hood of a car, Gustafson was in his basement revving up a bevy of experiments.

His parents gave him space in the basement of their Des Moines home to set up a small chemistry lab. The lab soon took over the basement.

"When my father [a pediatrician] would get rid of an autoclave, I'd add it to my collection," he said. "I had a centrifuge and a helium-neon laser. I was making holograms and having a good old time."

At 16, Gustafson built a harpsichord which he still has today. His motivation for all of these projects was simple.

"I wanted to learn as much about everything as possible," he said. Once in college, he focused on getting degrees in applied math because it offered the broadest science education possible.

After earning his doctorate from Iowa State, Gustafson entered the world of computing in the early 1980s. He was entering the computer industry just when the notion of parallel processing, a form of turbo-boosting the speed and power of computers by dividing problems and synchronizing the workload of processors, was being explored. Much to his dismay, industry people did not see a future in parallel processing.

"People had devoted their lives to conventional processing and they thought parallel processing was a real threat," he said. "People were saying, 'This will never work and we hope it never does.'"

After a couple of jobs in industry making and helping sell high-performance computers, Gustafson wanted to move to the other side -- to use the super-fast computers. While at Sandia National Laboratories, a U.S. Department of Energy lab in Albuquerque, Gustafson worked with computer scientists who set out to prove parallel machines could deliver on their promise of dramatically improving the power of computing.

He helped demonstrate a 1,000-times performance increase in a parallel computer over a conventional machine, a feat touted in Time, Newsweek, New York Times and the Guinness Book of World Records.

Finding his funding would follow him, he decided to return to Iowa and help set up a fledgling high performance computer lab at Ames Lab, now called the Scalable Computing Lab. Gustafson began with the idea of helping scientists use computers and to anticipate future computing problems before they become too unruly.

"Before the lab started, we had no money, no people, no equipment and no space. Otherwise we were all set," Gustafson said. But like the lab in his parent's basement, SCL began to take off.

Iowa State homepage

Inside Iowa State,, University Relations
Copyright © 1997, Iowa State University, all rights reserved
URL: html
Revised 10/23/97