CMS Bulletin

COMPUTING

Introduction

Computing activity has been running at a sustained, high rate as we collect data at high luminosity, process simulation, and begin to process the parked data. The system is functional, though a number of improvements are planned during LS1. Many of the changes will impact users, we hope only in positive ways. We are trying to improve the distributed analysis tools as well as the ability to access more data samples more transparently. 

Operations Office

Figure 2: Number of events per month, for 2012

Since the June CMS Week, Computing Operations teams successfully completed data re-reconstruction passes and finished the CMSSW_53X MC campaign with over three billion events available in AOD format. Recorded data was successfully processed in parallel, exceeding 1.2 billion raw physics events per month for the first time in October 2012 due to the increase in data-parking rate. In parallel, large efforts were dedicated to WMAgent development and integration with emphasis on increasing flexibility and reducing production tails. Keeping the sites operational and the central services up and running required constant attention and care. The transition to GlideIn WMS submission is progressing and the deployment of new technologies like CVMFS for the software distribution and the global CMS xrootd federation will help in the future to increase productivity of the computing infrastructure. This work will not stop during LS1. After the final processing of 2012 data, we will concentrate on upgrade MC and 13TeV MC campaigns and all the planned improvements and upgrades to the computing infrastructure.

Figure 3: Re-reco – number of events per month in 2012

Figure 4: Monte Carlo – number of events per month in 2012

Physics Support

The team effort followed a two-pronged approach: from one side with education, documentation and training activities; from the other by providing real-time help and troubleshooting on using the distributed computing system. Focus has been on stabilising the CrabServer operations and on introducing a new scheduler (remoteGlideIn) aimed at simplifying and improving Crab2 for both users and support.

The first ever CMS Data Analysis School (CMSDAS) was held in Asia at National Taiwan University, Taipei in September. The goal of having 100% Asian students was reached. The next CMSDAS schools are being held at the LHC Physics Center (LPC) at Fermilab (8–12 January 2013) and Terascale at DESY (14–18 January 2013). A Physics Analysis Tools (PAT) tutorial was held at CERN during 3–7 December 2012. A survey regarding "CMS Analysis Practices" will be announced shortly for feedback.

Computing Integration

The present gLite middleware components are reaching their end of support and will be developed further by the European Middleware Initiative (EMI) project. Therefore, a WLCG effort is ongoing to establish the new middleware flavours available at the Grid sites. The CMS integration team is participating by running tests using the CMS-specific submission tools and applications at selected sites that already provide the new middleware. Special emphasis is put on the validation against the various storage techniques used at the sites. Problems that were found during the testing either got fixed or were provided with workaround procedures.

Computing Evolution and Upgrade

The resources of the CMS online farm (about 3000 computers) will be available during LS1 for offline work. An Open Stack cloud layer has been deployed on a part of the cluster as a minimal overlay so as to leave the primary role of the computers untouched while allowing an opportunistic usage of the cluster. Jobs will be submitted to the system by the standard computing tools via a GlideIn WMS. Data present on the Tier-0 and the CAF will be accessible thanks to an xrootd cache server. The system will be fully operational during the first months of 2013.

Computing Resource Management Office

The Mid-2012 Resource Utilization Report and the Extra-2013 Resource Request Report was submitted to the Computing Resource Scrutiny Group. Feedback was provided at the RRB Meeting on 30 October: it was generally positive, in particular since none of the approved resource requests have a deficit beyond 10% compared to the Extra CMS request. The official 2013 pledges provided by participating countries/sites show a deficit of 14% in Tier-1 CPU compared to our needs, which will be partially compensated by the utilisation by Computing of the CMS HLT Farm during the LS1. One very positive note is an excess of +11% of pledged Tier-2 disk resources compared to the original CMS request.

The Resource Planning Spreadsheet is being updated, in view of the feedback received and in particular in terms of data-management plans and the 2013 re-processing plans. ESP approvals for 2012 are being finalised and a review for 2013 crediting policy is on-going. It has been agreed for the CMS Computing Shifts during LS1 to continue with the Computing Run Coordination (CRC), but without the requirement to be located at CERN, and with the Computing Shift Personnel (CSP) procedures, but with reduced manpower (–1/3). Consequently, a 2013 CSP Planning process was launched in collaboration with all participating institutes.