Skip to main content

Computers and tanks

ORNL's Titan supercomputer.

The waste tanks at Hanford and Savannah River may not be DOE’s only environmental challenge, but they’re at the top of the list. 

Hundreds of tanks hold 90 million gallons of highly radioactive, extremely toxic liquids and sludges. The tanks are in constant flux, although they are nowhere near evenly mixed. Not only do we not know what to do with many of them; we’re also still working to figure out exactly what’s in there.

Supercomputing may help us better understand the problem and eventually solve it. 

ORNL is home both to the world’s second most powerful supercomputer, Titan, and to deep expertise in data analysis. Indeed, supercomputing is the only tool that can take data at wildly different scales—from quintillionths of a second to hundreds of years, say, or from individual atoms to areas the size of a refrigerator—and tie them together.

“Can we model all of those complexities to the level of accuracy where it would be useful for cleanup and risk assessment? Maybe not,” said Bobby Sumpter, a computational chemical physicist at ORNL. “It is an area ripe for forefront-pushing. We’re beginning to realize that we do have a lot of data and we do have models. We should be able to make each of them better.”

It’s an approach made possible by the phenomenal growth in computing power, he said. Today’s most powerful computers are more than 100 times faster than those of a decade ago and more than 100,000 times faster than those of a decade before. Even so, Sumpter said, the problem will ultimately require machines dozens of times more powerful than today’s top systems: exascale machines capable of more than a billion billion calculations each second.

“This is an exascale computing problem,” he said. “It’s not that you have an exascale computer and solve the problem. It’s that you have an exascale computer, and you get the pieces of the models, and you incorporate the data you have to collect. It’s a major effort.”

Still, he said, such an effort would have been unthinkable not so long ago. 

“They’re trying to get this from where it is now to where it needs to be, realizing that we do have computer power to support it. Ten years ago, there was no chance that we could do that; the computer power wasn’t there.”

See also: