CMS Bulletin

COMPUTING

Computing activity had ramped down after the completion of the reprocessing of the 2012 data and parked data, but is increasing with new simulation samples for analysis and upgrade studies. Much of the Computing effort is currently involved in activities to improve the computing system in preparation for 2015.

Operations Office

Since the beginning of 2013, the Computing Operations team successfully re-processed the 2012 data in record time, not only by using opportunistic resources like the San Diego Supercomputer Center which was accessible, to re-process the primary datasets HTMHT and MultiJet in Run2012D much earlier than planned. The Heavy-Ion data-taking period was successfully concluded in February collecting almost 500 T.

Figure 3: Number of events per month (data)

In LS1, our emphasis is to increase efficiency and flexibility of the infrastructure and operation. Computing Operations is working on separating disk and tape at the Tier-1 sites and the full implementation of the xrootd federation to enable processing, production and also analysis more independent of the location of the input samples. The global GlideIn WMS pool will contribute to the efficiency increase as well and allow for more fine-grained prioritisations of jobs, from which also analysis jobs will benefit.

Figure 4: Number of events per month (Monte Carlo)

Figure 4: Number of events per month (Monte Carlo)

For the MC production, the bulk of 8 TeV MC samples have been completed and ramped up again in May with the upgrade samples and additional 7 and 8 TeV requests.

Many members of the Computing Operations team moved on to newer endeavors. Christoph Wissing joined the Computing Operations coordination team and, after four years leading Computing Operations, Markus Klute was succeeded by Christoph Paus.

Physics Support

As part of CMS-wide effort to bring new CMS users up to speed in doing physics analyses, Physics Support recently held two CMS Data Analysis Schools, one at the LPC in Fermilab (8–12 January) and other at the DESY (14–18 January). Over 200 participants took part in the hands-on physics analysis sessions. In addition, there is an upcoming PAT tutorial at CERN (15–19 July) limited to 25 registrants.

Besides smooth and stable running of analysis jobs across Tier-2 centres, during the past year there has been a steady increase in the number of job slots used for CMS analyses, reaching a level of ~26,000 concurrently running analysis jobs on average.

Figure 5: Analysis job slots used per week at T2 sites

Computing Integration

Following a common effort together with ATLAS to incorporate common tools in the distributed computing system, the Computing Integration group is participating in the commissioning and deployment of CRAB3, a new workflow management system for analysis jobs. CRAB3 incorporates new features like asynchronous stage-out of user data, automatic publication in DBS, better monitoring etc. The current CRAB3 prototype is based on the PanDA workflow management system used by ATLAS for production and analysis. A complete PanDA infrastructure is being setup at CERN for CMS by the Integration group in collaboration with CMS analysis operations and ATLAS and CMS developers.

CMS is committed to moving forward in incorporating multicore processing and scheduling into its computing system. The CMSSW framework is incorporating support for multithreaded data processing. The planned strategy is to schedule in the distributed computing infrastructure multicore pilots, which will manage the allocated cores. Multicore pilots can schedule single and multicore applications simultaneously on the same slot. The Computing integration group is performing tests to commission the multicore scheduling infrastructure.

CMS began to incorporate opportunistic resources into its computing system. At the moment we are still working on a seamless integration into the CMS production submission infrastructure (GlideIn WMS), but we already managed to run at limited scale and are in the process of scaling up.

Computing Resource Management Office

The RMO is active in finding and recruiting MnO-A manpower for the coming two years.