Many OSG applications have now been tested on the software release candidate for OSG 0.8.0 and have helped ensure that the release is as robust as possible. We have had an active group of more than 25 resource administrators and users participating in testing.
The last stage of the release process has been completing documentation for administrators and users of the software. The goal is to provide a comprehensive set of reference documentation for installation and configuration of each component of the software release.
The release documentation is available at its initial location and is being developed as part of the integration activity. It will eventually be migrated to the 0.8.0 release documentation when ready.
We are further developing the documentation hub which provides an entry point for any OSG member or visitor to all the documentation available on the OSG Twiki/work site.
We appreciate the contributions of the OSG members who are contributing to the writing and review of this documentation. And we welcome further contributions and comments from anyone.
~ Rob Gardner, Alain Roy & Anne Heavey |
| If you've ever flirted with the idea of contributing to International Science Grid This Week , now would be a great time. In the last four months iSGTW readership has more than doubled, indicated increased interest in grid computing projects and a great chance to increase your project's visibility.
iSGTW is a collaboration between the Open Science Grid and Enabling Grids for E-sciencE projects and promotes the success of grid computing as a tool for scientists and researchers.
The newsletter welcomes contributions in the form of project profiles, opinion pieces, technology overviews and more. Contribution guidelines are available from the website: http://www.isgtw.org
Subscription to iSGTW is free. |
|
|
Simulation of a detection of the Higgs boson in the CMS experiment.(Click for larger image.) Image Courtesy CERN |
The CMS-wide CMS Software and Analysis Challenge 2007 started about one month ago. In the US, the OSG Tier-1 and Tier-2 farms and storage resources are used. The goal of the challenge is to exercise the full CMS system at more than 50% of the scale needed for full data taking next year and to stress the system with “real users.”
The challenge has successfully met the management of seven physics streams at a total transfer rate of 300 MB/sec from CERN to the global Tier-1 sites and 200 MB/sec from the Fermilab Tier-1 to the 9 OSG Tier-2 sites. The next step is for physics groups to run analysis jobs at the Tier-2s and stimulate data placement from the Tier-1 data locale to the Tier-2s.
In all, more than 200 Tier-1 to Tier-2 permutations are exercised. The CSA07 goal is 100kjobs/day successfully submitted across the CMS sites. US CMS continues to solve problems making the overall system more robust. There are more than 20 million reconstructed events - about 50% of the goal (100 Hz, 24x7) and a rate of 17k analysis jobs per day, roughly 25% of the target.
~ The US CMS Team
|
| osg-vo-tests is an extensible Perl package which enables application owners (people responsible for running an application on OSG) to characterize OSG sites from the point of view of being able to run their application.
The motivation behind this package is to give application owners more information than VORS. While an invaluable tool within its scope, VORS does not address:
- Whether my jobs will be authorized and run (the dreaded GRAM errors 7, 47 & 22).
- Whether grid-ftp works from a worker node; SQUID? srmcp?
- Does MPI work? What is the compile command?
- Other application libraries and utilities: DB clients, Curl, Ruby?
- How many jobs are running right now? How many slots are free?
- Is OSG_APP writable from the WN? OSG_DATA?
- Will a batch job of mine ever actually start?
In addition, since there are multiple ways of discovering the information an application owner needs (fork jobs, VORS, ReSS ClassAds, batch jobs), there is difficulty collating, presenting or using information from these disparate sources. How can one tailor question sets for multiple combinations of requirements for different applications? osg-vo-tests attempts to fulfill these needs.
osg-vo-tests is written in Perl, for several reasons: extensibility through module inheritance and dynamic code evaluation; there is a good chance of familiarity for VO application owners; and one can have a fast development / test cycle.
In addition, there are several other attractive features of this package:
• A human-readable HTML summary page is generated with links to detailed test results; also CSV for machine-readability.
• The package obtains VO credentials (a VOMS proxy) before running the tests.
• Batch job submission and monitoring are handled, including a time-out facility.
• Application owners can add new tests and produce canned test list configurations for different applications.
Example test run results are available for the Engage and OSG VOs. Full documentation, download instructions and example configurations are available on the package's home page.
~ Chris Green |
|
|
| I presented a poster representing OSG at the third annual e-Social Science conference in Ann Arbor in early October. Below, I share some of the ideas and issues this community brings to Cyberinfrastructure.
The social science community sees CI as a social endeavor as much as a technical one. Issues of ownership -- trust, law, standards, privacy, cooperation, competition and assumed knowledge levels -- are as important to the success of an infrastructure as the technical implementations. This community is in the unique position of being able to both benefit from CI to advance its own research and to study and inform the sociology of CI projects.
Uptake and adoption of CI is seen as a complex social process. Barriers to adoption identified by Zack Kertcher and his colleagues in Accelerating Transition to Virtual Research Organization in Social Science (AVROSS): First Results from a Survey of e-Infrastructure Adopters include:
- Gaps in CI documentation, training and support for the non-computer trained
- Inadequate (CI-related) leadership, funding and qualified staff within the domain community
- Failure to keep users involved in all stages and to find ‘champions' among domain scientists.
- Failure to reconcile the agendas of computer scientists and domain scientists
The e-Infrastructure items used most frequently (so far) in social science are communication and collaboration tools, distributed data storage, and high bandwidth networks.
There seems to be agreement that “data is where the rubber hits the road.” Social science data comes in all shapes and sizes. To plan for data handling, it's critical to understand what is guiding the current end-user's practical purpose, while still allowing for different studies and scales, and for unforeseen future use. Can methods transfer from one data type to another? The area of data sharing needs lots of study.
How did OSG fare at the conference?
The keynote speakers identified TeraGrid as the CI “workhorse” available to their listeners, and failed to mention OSG. On the bright side, about a dozen people visited our poster, asked questions, and took brochures. Some were interested in resources, others approached us as an object of study.
~ Anne Heavey |
|
