The 2005 Summer Workshop. |
The third Grid Summer Workshop will be held June 26-30 on South Padre Island, Texas.
The OSG is a sponsor of the week-long workshop, in which approximately 40 undergraduate
and graduate students learn about distributed computing and its application in scientific
data analysis.
Last year's students hailed from 23 universities and four countries, and represented fields
from computer science and engineering to physics and atmospheric science.
The team that develops the curriculum, hands-on exercises and network design
for the workshop includes members of the GOC and the VDT team, OSG site
administrators and the University of Texas at Brownsville outreach group.
The workshop provides a basic foundation in
distributed computing and hands-on training in distributed computing techniques,
introducing essential skills that will be needed by students to conduct and
support large-scale scientific analysis in the emerging grid computing environment.
If you are interested in helping with this year's workshop, please contact me
for more information or visit the Web
site.
The workshop is also sponsored by the Center for Gravitational Wave Astronomy, a NASA
University Research Center at the University of Texas at Brownsville, iVDGL and GriPhyN.
Mike Wilde, OSG Education, Outreach and Training Coordinator
|
  |
|
 |
|
Throughout this past month there has been a lot of work on the configuration, management and usability of OSG 0.4.1, in parallel with application validation of the release. I encourage you to read and comment on the 0.4.1 documentation and give us feedback on your experience finding and using the information.
The Council is currently gathering detailed information about site and VO metrics. We want to better understand the large gap between the number of CPUs declared to be accessible to grid jobs and the number of jobs that have executed. We are also looking for ways to gather information about scientific results that have used the OSG.
GADU is now sustaining the use of more than 600 CPUs and is working with the Applications
Coordinators and GOC to double this number. ATLAS is the largest steady-state user with
almost 1,000 jobs running continuously and CDF is sustaining around 500 jobs when they are
submitting work. CMS reports steady progress in sustaining robust user job submission to
sites on EGEE and OSG using the WLCG Resource Broker as a single submit point. The operations
team is working to provide better information to VOs about the sites that support them, and
on helping sites with configuration and support.
Site administrators - we need your help! We need an additional person to work on and report
on improving the effectiveness of cluster configurations, to gather and distribute monitoring
scripts and to test proposed changes. If you are interested, please contact me.
Ruth Pordes, OSG Executive Director
|
Jeff Linderoth |
The Grid Laboratory of Wisconsin (GLOW) supports many different
scientific applications at the University of Wisconsin-Madison,
running the applications on local university resources. Dan Bradley,
of the Data Intensive Science University Network (DISUN) project at
Wisconsin, developed services to route GLOW jobs to OSG sites in
cases where "bare" compute and storage elements are sufficient. In
April this service has been put into production for a mathematical
research application I developed in collaboration with Francois
Margot from Carnegie Mellon University and Greg Thain from the
University of Wisconsin-Madison.
The application is designed to solve the football pool problem, one
of the most famous open problems in coding theory. The goal of the
problem, which gets its name from a lottery-type game where
participants predict the outcome of soccer matches, is to determine
the smallest "covering code" of ternary words of length six. (Or in
the football pool, to determine how many lottery tickets one would
have to buy to guarantee that no more than one prediction is
incorrect.) Even after decades of study, only fairly weak bounds are
known on this value. Solutions to this problem have applications in
data compression, coding theory and statistical designs. An integer
program to solve the problem is easy to formulate but extremely
difficult to solve.
We have previously run the application using resources from a local
Condor pool, GLOW and DISUN resources, and TeraGrid. Once we
decided to route GLOW jobs to OSG, the jobs were able to ramp up on
the remaining DISUN sites on OSG and to a further three OSG sites
within one day and have run continuously since then. The team
developed a specialized scheduler that carries out a job routing
policy that transforms excess jobs from Condor's vanilla universe
into grid universe jobs, directed to sites on the OSG. Ideal jobs for
such routing should have all input/output files explicitly specified
and Condor's file-transfer mode enabled.
These diverse resources have helped us make progress on the football
pool problem, and a full run of the application is planned for later
this summer.
Jeff Linderoth, Lehigh University
|
|
|
