GlueX Computing Effort Gets Nod from NSF

A grant from the National Science Foundation is helping researchers to gear up for GlueX data analysis, long before the experiment even begins....

To help improve our understanding of the universe, more than 185,000 computer owners in 202 countries have volunteered their computers' spare time to the Einstein@home project. The project splits huge amounts of data from giant telescope arrays into small chunks that home computers can analyze. This allows researchers to look for wrinkles in space known as gravitational waves. Now a Jefferson Lab-based group is taking this idea, called "distributed computing," and applying it to meet the demands of particle physics.

The Gluonic Excitations Experiment (GlueX), set to run in Jefferson Lab's Hall D after the 12 GeV Upgrade, is expected to generate huge amounts of data – about 2 Gigabytes per second. Scientists will need distributed computing to make sense of it all. But instead of ordinary computers – they’re planning to split the analysis among cluster-style supercomputers.

Several of the universities collaborating on the GlueX experiment have received a grant from the National Science Foundation to develop and test what would be the next generation of data analysis tools. Once complete, scientists would be able to distribute processing simultaneously over a cross-country network of cluster computers.

"NSF is funding three-quarters of a million dollars to engage people in an interesting computing project that's going to push the edge of data analysis and help us do better science in the future," says Matthew Shepherd, an assistant professor at Indiana University and co-principal investigator on the grant.

"It's a proposal funded through the NSF program known as Physics at the Information Frontier. It funds projects that are at the intersection of science and computing. Our proposal is to help develop an algorithm for data analysis on data that will come as part of the 12 GeV Upgrade at JLab," he says.

Scientists in the GlueX collaboration will study "glue". Not Elmer's or Loctite – but the glue that binds the foundations of matter that make up our universe. This glue binds together the bits of matter that build the protons and neutrons in every atom.

It’s not possible to scoop a glue sample out of a proton and plop it down on a microscope for examination. But scientists think they can study this glue by producing a new type of particle called an exotic meson. This particle has the same building blocks as a proton: quarks. Welding the two quarks together in an exotic meson is a "flux tube" of glue – the kind the scientists want to study.

Unfortunately, signatures of these exotic mesons will be buried in a mountain of data from other particles that are also likely to be produced in the GlueX experiment.

"If you were to do this type of analysis on your personal computer, there's no way that you could do this in any reasonable amount of time," Shepherd explains. "An analysis the Indiana University group did about two years ago consumed 10-to-20,000 hours of computing time. And for GlueX, we're going to have data sets that are maybe 10 or 100 times that. This presents a challenge for analyzing the signatures of exotic mesons that we are tackling now, five years before the data arrives. With these NSF funds, we will make great advances in the speed and efficiency of the data analysis. "

The proposal funds three postdoctoral research associates to work with scientists at Indiana University, Carnegie Mellon, the University of Connecticut and Old Dominion University. These physicists are developing data analysis tools.

The tools are for open access amplitude analysis on a grid, a fancy term that means the tools will, simply put, allow researchers to use supercomputers connected via the Internet to analyze particle physics data in concert.

Ultimately, Shepherd would like to make it possible for any nuclear physicist or graduate student to use the tools developed for GlueX to analyze data, with minor tweaks, for almost any particle physics experiment.

"What we really want are off-the-shelf tools that anyone can take and use right away, where we've solved all the hard computing problems. You can think about physics and not about programming and interfacing multiple computers and networking and all that sort of stuff," he says.

By Kandice Carter
JLab Science Writer.