|A Big Step Forward for the TeraGrid|
Anglo-American Team Gets Big Results with TeraGrid
Linking resources on two continents allowed The TeraGyroid Project to do a year's work in three months.
PITTSBURGH -- In a three-month project centered on the Supercomputing 2003 conference last November in Phoenix, a Boston-U.K. team of scientists linked more than 6,000 processors and 17 teraflops of computing at six different facilities on two continents. Their Grid-based effort has led to significant new scientific understanding and represents probably the most impressive example to date, say scientists, of how Grid computing gives a powerful boost to large-scale scientific computation.
"The Grid creates a tremendously powerful environment," said University of London chemist Peter Coveney, who led the U.K. side of the TeraGyroid Project. With a sophisticated approach called "computational steering" the researchers launched multiple simulations from which they chose a set of dynamics on which to focus at large-scale. As a result, said Coveney, "our productivity skyrocketed."
"Because of the Grid," said Bruce Boghosian of Tufts University, "we've been able to address a large-scale problem of genuine scientific interest and dramatically reduce the time to insight. We've made progress in three months that would have taken more than a year by conventional methods." Boghosian, a professor of mathematics, leads the stateside contingent of the TeraGyroid effort.
For the November SC2003 effort, prepared in advance by two months of computing in the U.K. and at TeraGrid sites, the researchers linked the National Science Foundation's TeraGrid with the U.K. E-Science Grid via dedicated trans-Atlantic fiber. They tied together resources at Daresbury Lab, U.K. (IBM Power 4 Regatta) and Manchester, U.K. (SGI Origin 3800 & SGI Altix) and at four TeraGrid sites: Pittsburgh Supercomputing Center (HP-Compaq TCS-1), the National Center for Supercomputing Applications (Itanium 2), San Diego Supercomputing Center (Itanium 2) and Argonne National Laboratory (visualization cluster).
Using the NSF Terascale Computing System (LeMieux) at Pittsburgh, they carried out the largest simulation of its type (lattice-Boltzmann method) to date, with resolution of more than a billion lattice-sites. Their work focused on complex materials shapes, known as gyroids, with properties in between liquid and solid.
With follow-up computations into February 2004, they produced more than three terabytes of useful data. To gather and collate this quantity of data from multiple sites was a large task in itself, and the team is still analyzing results, which point to new understanding of the liquid-crystalline materials they study, with several papers in preparation.
At SC2003, the TeraGyroid Project was recognized as the "Most Innovative Data-Intensive Application." More recently, the project received a 2004 ISC Award, the major supercomputing award in Europe, for "Integrated Data and Information Management."
"The ability to explore over a billion lattice sites could be accomplished only by sharing the processing horsepower at the U.K. and U.S. sites over a transatlantic grid," said José Muñoz, senior scientist at NSF's Division of Shared Cyberinfrastructure. "The goal of the TeraGrid is to enable exactly this sort of new science and the related scientific breakthroughs, and we expect many innovative experiments to follow in the footsteps of the TeraGyroid effort."
"The TeraGyroid Project exemplifies what's possible with Grid technologies," said Rick Stevens of Argonne National Laboratory and the University of Chicago, TeraGrid project director. "It's a major success for the NSF vision of integrated national cyberinfrastructure, and it helps us to appreciate that -- just as the economy is global -- with the Grid, science too has become global."
Boghosian and Coveney emphasize that the vision of Grid computing -- "transparent" access to resources without regard to location -- is not yet reality, and it was a large task to coordinate this ambitious project, which involved about 30 organizations and more than 100 individuals.
The TeraGrid is a multi-year effort to build and deploy the world's largest, most comprehensive distributed infrastructure for open scientific research. The TeraGrid also offers storage, visualization, database, and data collection capabilities. Hardware at multiple sites across the country is networked through a 40-gigabit per second backplane-the fastest research network on the planet.
The TeraGrid sites are: Argonne National Laboratory; the Center for Advanced Computing Research (CACR) at the California Institute of Technology; Indiana University; the National Center for Supercomputing Applications (NCSA) at the University of Illinois, Urbana-Champaign; Oak Ridge National Laboratory; the Pittsburgh Supercomputing Center (PSC); Purdue University; the San Diego Supercomputer Center (SDSC) at the University of California, San Diego; and the Texas Advanced Computing Center at The University of Texas at Austin.
More information, including graphics:
A report, "The TeraGyroid Experiment," by participating scientists:
The TeraGyroid Project was funded in part by the UK's EPSRC and used capabilities developed in the RealityGrid project: