E-Science: Massive experiments, global networks

Although access to the data is restricted to physicists in countries collaborating on the collider, the experiment’s global network taps the number-crunching ability of more than 140 computer centers in 33 countries. Much of the grid is designed to distribute the data for analysis locally, notes Ken Bloom, a physicist at the University of Nebraska and a member of one of CERN’s collision-detector teams.

“In the good old days, you’d have one biggish computer at a lab doing all the data processing,” he says. But the volume of data expected to pour from the collider’s detectors is so large that a single-center arrangement wouldn’t work – the power and cooling requirements to keep the computers happy would be enormous. As a result, the data center at CERN makes only a first pass at any processing, then it archives the data and sends subsets out to a second layer of computing centers. Physicists enter the picture only after the information is further subdivided and sent speeding across the Internet to two more layers of computer centers – the last of which are at individual universities.

It’s there that the discoveries will emerge in countries ranging from Japan, the US, and European nations to Pakistan, Turkey, and Brazil.
“We’re doing what we can to level the playing field for people who are collaborators,” giving them equal access to the data, Dr. Bloom says.

By contrast, the US’s TeraGrid gives any university researcher in the country access to an enormous amount of processing power – at no cost other than the time it takes to fill out an application for computing time. The grid’s workhorse computers are housed at nine sites across the country and are being used for everything from modeling severe weather to economic scenarios, notes John Towns, director of persistent infrastructure at the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign.

Ironically, he says, despite the network’s ability to help drive science and broaden access to high-powered computing, “many people don’t know of the TeraGrid’s existence.”

If they did, it still might be the wrong tool, observes Gerhard Klimeck, a professor of electrical and computer engineering at Purdue University and an assistant director for the Network for Computational Nanotechnology. Large networks like TeraGrid are well suited to high-octane computational tasks, he says. But in fields like nanotechnology, a lot of work is done at the lab bench, where researchers with data and a question want an answer in minutes – not the hours it takes to fill out and process applications.

To fill this need, nanoHub was born. “Our goal was to put nanotechnology simulation tools in the hands of people who otherwise wouldn’t touch them with a 10-foot pole,” Dr. Klimeck says. It provides access to sufficiently sophisticated computing power and programs to serve the “I need the answer yesterday” crowd.

NanoHub began in 2002, has six universities as its core, and has grown from roughly 1,000 users to some 77,000 a year, Klimeck says. Perhaps as important is the growing number of research papers that credit nanoHub for providing computational tools. NanoHub has since spun off clones at Purdue. Because it’s readily available to anyone who signs up for an account (no formal application for computer time necessary), nanoHub is also used for engineering education, he says.

“E-science is rapidly changing,” notes Indiana’s Dr. Fox. And with it comes the potential for an increasingly level playing field among researchers in the US and overseas, he says.

Previous

1 | 2