"The computer scientists wanted to get more people involved in grid computing, and many of the researchers on campus needed more resources," says Sridhara Dasu, a particle physicist and the technical lead for GLOW. "We hoped that by pooling our talents and resources, everyone would benefit."

And benefit they have. GLOW has been one of the largest producers of simulated particle physics events for the CMS experiment, last year creating 30 million events that will help predict and analyze the behavior of the newly-constructed particle detector. Millions of neutrinos passing through the massive IceCube detector at the South Pole have been simulated and analyzed. And more than 30 graduate students in chemical and biological engineering have published over 100 papers with simulations performed using GLOW resources.

"All the computational work we do uses GLOW," says chemical and biological engineering professor Juan de Pablo, one of GLOW's six principal investigators. "We simulate the behavior of different types of matter at the nanometer scale."

It's extremely difficult to measure the properties of the materials de Pablo's group investigates. At tens of nanometers thick—a human hair is 1,000 times wider—the behaviors of very thin films can vary widely from those of larger amounts of the same material. Researchers use computer models to predict the behavior of these films, using experiments when possible to test their models. De Pablo's group studies polymer coatings on computer chips, the behavior of long DNA strands in nanofluidic devices, and the structures and functions of nanometer-scale biological membranes.

"We simulate how the behavior of cell membranes is altered when various additives are used," explains de Pablo. "For the additives to be useful for therapeutic purposes, we need to know how flexible and porous a membrane is, what its composition is, and how all these factors influence the behavior of a cell."

GLOW was established through a grant from the National Science Foundation and support from the University, with ongoing operation led by the Condor project and supported by the NSF Middleware Initiative. Each of GLOW's participating research groups received and configured several racks of computers, where they receive highest priority, with any unused computing cycles shared equally among all groups. As new members have joined and existing members have added resources, GLOW has increased to more than 1,000 CPUs. Researchers now work to interface GLOW with other grids, including the Open Science Grid and Harvard's campus grid, the Crimson Grid.

"We've recently added users from the ATLAS particle physics experiment and condensed matter physics, and there is interest from many other groups," says Dasu. "We've even contributed indirectly to research at the university—several social scientists have studied the GLOW phenomenon."

Learn more at the GLOW Web site.