In January, almost 300 members of the Worldwide LHC Computing Grid (WLCG) collaboration attended a week-long workshop at CERN to discuss the status of the infrastructure, as well as detailed plans and timescales to prepare for the start-up of the LHC later this year. The week included experiment-specific sessions as well as a joint-operations workshop, and followed a dCache workshop at DESY the week before. While previous workshops relied mainly on presentations, the January event was more interactive and included various informal "birds of a feather" sessions to discuss topics of shared interest, such as data management, monitoring and user support.
The WLCG was formed by resource providers – Grid projects, mainly EGEE in Europe and OSG in the US, and individual resource providers – to deal with the 15 PB of LHC data expected every year. The computing sites are arranged in a number of tiers, with CERN serving as the Tier-0 site, which will collect and distribute data to 12 Tier-1 sites. Some 150 Tier-2 sites will help process the data. The CERN workshop was the first to address the full WLCG team and brought together people from 27 countries and 86 sites.
All four large LHC experiments organized sessions to allow direct contact between site managers and experiments experts. These were well attended and the participants judged them useful. The ALICE session concentrated on different tutorials regarding specific aspects of ALICE software such as monitoring, AliRoot and AliEn. Topics in the ATLAS session included data management, storage-resource management and the security model of the services deployed, while the CMS session covered file-transfer and integration plans as well as computing resources and storage classes. Discussions in the LHCb session included topics such as testing of the "glexec" middleware module by some sites to permit various levels of access to computing elements and worker nodes, data security and data transfer between sites.
The operations workshop concluded that as Grid operations are maturing, preparation for the LHC start-up is the main driving force behind infrastructure changes. Several issues, such as portability of the gLite Grid middleware and migration to the Scientific Linux 4 platform and 64 bit support, still have to be addressed. These and other changes, however, will be introduced only if they are not disruptive, so as to ensure a ready, reliable and stable service to the LHC experiments.
Three working groups have been set up to focus on improving service and site reliability, which are all coordinated. The Grid Monitoring Group will pull together monitoring data and provide views for the different stakeholders. The Site Management group will work to harmonize tools and best practices and will issue recommendations to improve site management. The System Analysis group will continue work done by ARDA (CERN Courier Jan/Feb 2005 p19) to provide feedback from the applications point-of-view. Another area still under development is the interoperation between the EGEE and OSG infrastructures.
Over the course of the year, the WLCG will continue to test computing models and basic services, in particular the full data flow from the trigger systems used by the LHC experiments through to distributing the data and performing analysis. Dress rehearsals over the summer for end-to-end tests of all components should bring the service to full capacity and performance, ensuring reliability and 24/7 operation, well in time for the LHC pilot run in November.
The next full WLCG collaboration week is planned for spring 2008, but this will be complemented by smaller, focused events such as the Operations Workshop on 11–15 June in Stockholm and meetings at events such as CHEP'07 on 1–2 September in Victoria, BC, Canada.
• For more information, videos and the presentations see http://indico.cern.ch/conferenceDisplay.py?confId=3738.