Share |

Feature - Coming: Sustained petascale computing for open science

Feature - Coming: Sustained petascale computing for open science


Illinois' Petascale Computing Facility (PCF), which will house Blue Waters, is planned as a leading edge, energy-efficient facility capable of housing multiple Blue Waters-scale systems.  The PCF will provide up to 20,000 square feet (about 1900 m2)and 24 MW for computer equipment (not counting cooling infrastructure) and is  expandable  to two to four times that.

Some of the innovative, energy-saving technologies in the building include 480V power (saves 5-10% in transmission loss), liquid cooling for computational components, and outside ambient air cooling for most of the year to avoid using artificially chilled water.

Image courtesy of NCSF.

Computing capacity has grown dramatically over the last decade, allowing scientists to tackle much larger and more complex research problems. The first sustained petascale computing system for “open” science — science done using open source tools — is expected to come fully online in 2011. Called Blue Waters, it will be capable of performing one quadrillion — that's one million billion — operations per second and will open up even greater opportunities to address some of most challenging problems in science and engineering.

Blue Waters is a joint project of the University of Illinois, its National Center for Supercomputing Applications (NCSA), the IBM Corporation, and the Great Lakes Consortium for Petascale Computation.  The system will contain more than 200,000 compute cores, about one petabyte of main memory, more than ten petabytes of user disk storage, and 500 petabytes of archival storage. Performance projections indicate that Blue Waters will deliver a sustained performance of at least one petaflop, one quadrillion calculations a second.

Comparisons to existing systems are difficult, says William Kramer of NCSA and deputy director of the Blue Waters project, since most metrics describe peak performance while Blue Waters is focused on sustained performance. He estimates that at peak, Blue Waters is between 6 and 10 times faster than any system existing today.  “But even this ratio is misleading,” he says, “since it focuses only on floating point operations.  In other aspects, Blue Waters is 10 to 50 times more powerful.”

This visualization from 2004 illustrates a high-resolution tornado simulation. When it comes online in 2011, Blue Waters will enable significantly more detailed simulations of many socially relevant phenomena, like the weather, new medicines and materials, and disease outbreaks.

In this visualization, orange spheres are rising; blue spheres are falling. Similarly colored streamtubes represent a 100-second history of selected weightless tracer particles. Swaying cones are used to symbolize the speed and direction of the wind at ground level.

There is a lot more information here.

Image courtesy of Bob Wilhelmson, Robert Patterson, Stuart Levy, Matt Hall, Alex Betts, and Donna Cox, NCSA; Lou Wicker, National Oceanic and Atmospheric Administration; and Matt Gilmore and Lee Cronce, University of Illinois at Urbana-Champaign.

To run on such a large number of cores and take full advantage of the system’s capabilities, researchers will need to modify their applications. Currently, few applications can use more than 10,000 cores, and according to Kramer, scaling up by a factor of ten or more could cause problems that prevent the application from running at peak performance.

“We are challenged to make sure Blue Waters works well for a diverse set of science fields, ranging from nanotechnology all the way up through astrophysics, climate change, chemistry and materials — any type of science you can envision,” says Kramer.

The National Science Foundation is hosting a series of international peer-reviewed competitions to allocate time on Blue Waters. To help prospective users prepare, the Blue Waters project is providing in-depth consulting services and application development workshops. In addition, selected groups with allocated time will participate in Petascale Application Collaboration Teams, where they will receive advanced help in optimizing and scaling their applications.

“The project involves many activities that will make it easier for scientists and engineers to make use of the Blue Waters system,” Kramer says. “We want them to be able to fully exploit Blue Waters’ unique capabilities and achieve the highest performance possible.”

Amelia Williamson, for iSGTW

No votes yet

Comments

Post new comment

By submitting this form, you accept the Mollom privacy policy.