Making history | National Center for Supercomputing Applications at the University of Illinois
09.24.12 - Permalink
by Barbara Jewett
Early results indicate Blue Waters will live up to expectations.
A brief but tantalizing test run of the Blue Waters sustained petascale supercomputer earlier this year has left researchers eagerly awaiting full deployment.
The Blue Waters Early Science System (BW-ESS) was only about 15 percent of the full machine as it consisted of 48 cabinets with 4,512 XE6 compute nodes and 96 service nodes and a Sonexion Lustre Storage Appliance provided two petabytes of disk storage. But that was enough to allow researchers to test the technology used in Blue Waters and identify and fix software bugs and other issues that prevented some codes from successfully scaling.
The results for many of the teams were astounding. Klaus Schulten, a biophysisist at the University of Illinois at Urbana-Champaign perhaps sums it up best in his report: "Not in our wildest dreams could we have imagined the greatness of the new NSF center machine. We are sure Blue Waters will make science and engineering history."
Schulten's team pursued three projects simultaneously. Their runs studying the HIV virus capsid yielded some unanticipated results. Enabled by the 2.9 release of NAMD, which includes a high-performance interface to the Cray Gemini network of Blue Waters, the Schulten team was able to explore aspects of the HIV capsid's structural properties in a 100-nanosecond simulation. This simulation revealed that the capsid is stabilized in a manner other than has sometimes been proposed. A manuscript is being prepared for publication detailing their findings.
The PRAC team led by Robert Sugar of the University of California, Santa Barbara, used their BW-ESS allocation to complete their calculation of the spectroscopy of charmonium, the positronium-like states of a charm quark and an anticharm quark. They did this using lattice gauge theory, an ab initio version of quantum chromodynamics (QCD) that can be simulated numerically. They focused on the experimentally well-measured low-lying states, but say in their report that the "ultimate goal is to use the same study to help characterize some of the mysterious excited states that could be molecular states of loosely bound charmed meaons or exotic states with a valence gluon in addition to the standard valence charm quark and anti-quark."
Their simulation produced a wealth of data with, they think, some very promising results, particularly the mass splittings of the 1S and 1P states. And the team is very pleased that with Blue Waters, statistical errors have been significantly reduced from previous studies. They presented their findings at the Lattice 2012 conference in Cairns, Australia, in June.
As reported in the summer issue of Access, the team led by Stan Woosley of the University of California Observatories assisted the Blue Waters visualization staff in refining user needs in addition to making some science runs. While it is generally accepted that the exploding star is a white dwarf driven to thermonuclear runaway by the accretion of mass from a binary companion, says Woosley, the nature of the progenitor star and how it ignites and burns are debated.
They accomplished their main goal of simulating on the BW-ESS a large fraction of the ~1 second it takes the bubble to reach the surface, using higher resolution than has been previously possible. In addition, they were also interested in how the background turbulence and choice of ignition location affect the bubble's evolution. So they performed five additional simulations. A somewhat surprising result of these additional simulations was that the background turbulence only seems to affect the flame morphology for ignition close to the center. The six simulations produced over 45 TB of data and the team is certain that further analysis will yield additional interesting scientific results.
Blue Waters is supported by the National Science Foundation through awards ACI-0725070 and ACI-1238993.