 |
A fundamental problem in high-energy physics is filtering through data. The LHC's detectors, for example, will track events that represent about one petabyte—or one quadrillion bytes—per second. A lion's share of these data are garden-variety events and particles that don't interest researchers. With its sensors honed to record only a very particular set of trigger events, the CMS detector will reduce this background by about seven orders of magnitude. Nonetheless, that leaves about 100 megabytes of data per second to digest.
A whopping 1.5 million CPU hours on the Wisconsin's Condor system, which pools the idle processing time of general-purpose Linux workstations for use in large-scale computations, are already being used by the Caltech team to simulate LHC events related to the CMS detector. The team is also using an HP-Convex Exemplar system at Caltech and Linux clusters at Caltech and CERN to run the simulations.
|
|
 |
|
Multi-jet event of the sort that will be detected by the Compact Muon Solenoid. Jets are conical sprays of subatomic particles confined to a certain narrow width. Image courtesy of CERN.
Click to enlarge the image.
|
|
|
"During peak processing times on Condor, we're submitting 600 jobs and have nearly 300 running," says Vladimir Litvin, a senior engineer in Caltech's high-energy physics program. "And the usual datafile from a production run is about one gigabyte. We need large amounts of processing power, storage space, and network bandwidth for our tasks."
The team has already simulated about 1.25 million LHC events related to the search for the Higgs boson, as well as another 1.25 million for other groups that will use the CMS detector to hunt other particles.
|
