SHAUNA STEPHENSON/WTE Aaron Andersen, a manager in the Scientific Computing Division, talks about the format of the supercomputer that is likely to come to Cheyenne. Andersen, who spoke to the WTE on Thursday at the National Center for Atmospheric Research in Boulder, Colo., said much of what is currently located in Boulder will be moved to Cheyenne when they relocate, but they will keep minimal operations running in Boulder.

Story by Jessica Lowell

rep5@wyomingnews.com

Photos by Shauna Stephenson

rep4@wyomingnews.com

There's no paper in Lawrence Buja's office.

On his big empty desk sits a big computer screen with a tiny keyboard attached and a smaller screen with a big keyboard. In the corner behind his door perch two espresso makers.

Buja is a climate scientist at the National Center for Atmospheric Research's Mesa Lab in Boulder.

His work takes up a lot of space and time, but it's not captured on pieces of paper tacked to his walls.

The bulk of his work and that of his colleagues is tied to supercomputers, like the one that's a floor below Buja's office with a view of Boulder Mountain Park - and the larger, more powerful one that is expected to be built outside of Cheyenne.

On his large computer screen, Buja runs a climate model. It's a program that collects and maps a wide range of information on weather, precipitation, volcanic activity and the like over centuries. It then generates a forecast of what's likely to happen to the climate on this planet in the future.

The final product also includes data from running a number of different scenarios to get a whole series of "what if?" pictures.

Making the model go took half of the capacity of the Boulder supercomputer for five months plus:

-One-third of the capacity of the supercomputer at the Oak Ridge National Lab in Tennessee.

-A substantial portion of the capacity of the National Energy Research Scientific Computing Center at the Oakland Scientific Facility in California.

-The services of a supercomputer in Japan.

That highlights two important things: The nature of NCAR is collaborative. And the work it does is not simple.

But as detailed as the model is, is could be better.

"With the new machine," Buja says, "we can start talking about smaller regions."

Right now the model is useful on a continental scale. But with more computational juice, projections could be made over an area that encompasses a smaller region such as eastern Wyoming, western Nebraska and part of South Dakota.

Understanding changes to the climate on that scale has practical value to policymakers, ranchers, farmers, the energy industry and a host of others.

"We can improve the realism of what is represented in the model," Buja says. "If we have more computing ability, we don't have to have inaccurate guesses."

Aaron Andersen's office overlooks only a conference room and a small reception area.

As computer operations manager at NCAR, Andersen oversees the equipment that makes the model that Buja worked on - and dozens of other projects - go.

The supercomputer itself is a series of computer banks joined by a network of cables under the raised floor to a switch that makes them operate as a single unit.

Row after row of black-encased cabinets do project work that's scheduled on a priority basis, much the same way that commercial air travel is organized.

Some projects go first class with no waiting, others go business class, and some are queued up in standby mode, going only when space opens up.

The soundtrack of this sophisticated computational work is the constant dull hum of machinery accompanied by a thermal element.

The air that the machines exhaust can reach 120 degrees, and that requires air conditioning to knock it down. In some cases, the conditioning is external, coming through vents in the floor. In other cases, the cabinet doors are water-cooled.

Running the computations is only part of the computing division's work. It also handles an enormous amount of data storage. It is dumped into cabinets at one of the supercomputer room on 200 gigabyte cartridges that are retrieved robotically on demand.

"We like to keep them separated," Andersen says, pointing to safety reasons. "We think 90 miles away is far enough."

Ninety miles is about the distance between the Mesa Lab and the Wyoming site that NCAR and University Corporation for Atmospheric Research officials picked as the site of the new supercomputer.

That site, 24 acres just west of Cheyenne, is expected to house the supercomputer for at least two decades after it's up and running in 2010 or 2011.

Supercomputer technology goes the same way that all technology tends to go: smaller and faster.

"It's Moore's Law," Andersen says. "You can double your performance just about every 18 months. You can get more stuff on a chip."

That drives changes in the supercomputer over time. NCAR puts its contracts for computers out for bid and typically signs a three-year contract with a two-year option to extend, so components are changed out regularly.

Scientists build their models with the constraints of existing technology, but they may want to add some other variable or calculation that requires a great deal more capacity than exists.

Even though the technology is getting smaller, the Mesa Lab is out of space for further expansion, and that's what drove the search for a new location.

"We accelerate faster because of the insatiable demand for computing ability," Anderson says.

When Buja and scientists like him want to refine their models and provide smaller-scale information, the models get built on smaller grids.

"Each time we reduce the size of the grid (on the model) by half, it requires four to eight times the computing ability, and that's not even adding all the layers on," Anderson says.

"The demand for new capacity gets eaten up right away."