Image Courtesy Mike Wilde |
You can view all the materials
from the 2006 Summer Grid Workshop which took
place June 26-30 on South Padre Island, Texas. The workshop was
co-sponsored by The Center for Gravitational Wave Astronomy (CGWA), the
Grid Physics Network (GriPhyN) and the International Virtual Data Grid
Laboratory (iVDGL).
The presenters included Gabrielle Allen, Rachana
Ananthakrishnan, Ben Clifford, Jaime Frey, Archit Kulshrestha, Jorge L.
Rodriguez, and Mike Wilde, who was the overall coordinator.
The 40 students included mechanical and bio
engineers from as far away as Uruguay and Puerto Rico. Some of the
students were able to convert their applications to run over the
TeraGrid and OSG sites that were available to the workshop participants
Mike Wilde, University of Chicago
|
|
|
We are all just back from the OSG Consortium meeting last week. Many
thanks to Gordon Watts and his team for hosting. There was much
interest in the plenaries -- especially the science presentations and
the work being done in campus and other grid organizations. There were
lively discussions at the parallel sessions and we solidified
requirements for the software releases in the next months -- now to
address priorities and the schedule. Mark your calendars for the next
face-to-face meeting at UCSD/SDSC either 5th or 12th March.
New Compute Elements from LeHigh University
and Notre Dame, and Storage Element from SPRACE in Brazil have joined
the OSG infrastructure. We continue to work to enable effective use of
OSG for LIGO, NanoHub and GLOW. Purdue, a new member of the OSG Council
since the February meeting, presented their Campus infrastructure and
goals of being part of the OSG.
In the past two weeks the overall job
throughput on OSG has increased by about 10% due to smaller VOs
increasing their testing of the infrastructure and one of the "big
three" CMS running production for the analysis challenge in the fall.
Ruth Pordes
|
Figure 1 (click to view larger image): A scan through matter anti-matter oscillation frequencies for
the Bs meson.The peak at 17.33ps^-1 leads to a measurement of the
oscillation frequency to within 1.4% |
CDF, with a 1.4PB dataset stored on  tape at Fermilab, is a physics experiment at the
forefront of the exploration of extensions to the "Standard Model" of
particle physics. Two examples of this research are illustrated in
Figures 1 and 2. Both of these results were presented at international
conferences this summer by the Collider Detector at Fermilab (CDF)
experiment. Both benefited from jobs running on the OSG and EGEE grid
infrastructures.
Figure 1 shows the precision measurement of
matter anti-matter oscillations in the Bs sector.
Figure 2 depicts a search for same-sign di-lepton events at high energy.
Both of these measurements impose stringent constrains on
extensions to the Standard Model, including Supersymmetry,
a theory that would explain the observation of dark matter in the Universe.
Figure 2 (click to view larger image): The leading lepton transverse momentum (pt) in the same-sign
di-lepton search. With more data, a hint of Supersymmetry might show
itself at large pt. |
There are typically 4000 to 5000 CDF
analysis jobs running at all times and about 600TB of data accessible
on disk, with roughly half locally at Fermilab and half at remote
clusters. Data is handled globally using SAM which accesses the
Fermilab tape archives via the dCache disk caching and Enstore mass
storage systems.
Today the majority of the CDF computing
resources outside Fermilab are accessible only via OSG or LCG
compute elements. To enable access for more than 700 CDF collaborators
worldwide, a group of UCSD and INFN physicists and computer scientists
modified the CDF job portal infrastructure to submit workloads
transparently through Grid interfaces. CDF is today one of the three
largest users of the OSG, with a peak of close to 1200 jobs running
simultaneously and a typical weekly consumption of 20,000 to 90,000
hours.
The hiding of the grid technology was
simplified dramatically by using condor glide-in technology. CDF had to
develop a "glide-in factory" that controls the glide-in submission to
the grids. Once a glide-in starts at an arbitrary grid site, it calls
home, and adds itself as an additional batch slot to a condor batch
cluster. This batch cluster fed by glide-ins looks no different to the
CDF portal than any of the dedicated condor pools at FNAL.
Frank Würthwein, University of
California San Diego
|
|
|
