Grid computing activities
Computing for the LHC experiments at HIP took a big step forwards largely due to the annual funding of 800 kEUR/year that the Ministry of Education has granted for the years 2008, 2009 and 2010. The funding is exploited according to the guidelines defined in the contract signed by HIP and CSC (IT Center for Science Ltd) and the work has been carried out in close collaboration between the HIP CMS and Technology programmes, CSC and Nordic DataGrid Facility (NDGF). This close collaboration resulted in several advances in many areas of CMS computing that are summarized here.

Hardware. The Sepeli computing cluster, situated in the CSC premises, was fully devoted to HIP for LHC computing usage as from mid-April 2008. A 170 TB disk system using the dCache software was connected to Sepeli for ALICE and CMS usage. The CMS CPU resources consists of Sepeli maintained by CSC and Ametisti maintained partly by HIP which have a total of 512 and 260 cores respectively. A new 10 Gb/s Optical Private Network (OPN) for Nordic LHC computing was created and one endpoint was created at CSC. The dCache system is able to saturate the 1 Gb/s interfaces it presently has to the dedicated LHC OPN. A hardware temperature monitoring tool for the Ametisti nodes was developed and installed allowing automatic shutdown in case of machine room cooling problems. A new shared Linux cluster, Korundi, was acquired together with the Departments of Chemistry and Physics. There are 400 cores on Korundi and 992 GB of RAM in total and 56 TB of raw disk. A 64-bit Rocks 5.1 cluster, tb64, was built using new and recycled hardware for development work on Korundi.

Software. The new manpower available due to the dedicated LHC computing funding enabled several major breakthroughs towards creating a commissioned Finnish CMS Tier-2 resource. NDGF and the Finnish national grid infrastructure M-grid uses the Advanced Resource Connector (ARC) middleware, so it was decided to continue the development of interfaces to the ARC middleware for CMS software services to benefit from the available expertise and common solutions. The problem of submitting CMS Remote Analysis Builder (CRAB) jobs to ARC Computing Elements (CEs) was solved by using and developing further the gLite-ARC interoperability work carried out during the last few years. HIP contributed to the CSA08 Analysis Commissioning in May-June 2008 by using gLite-ARC interoperability to run CRAB jobs submitted with both the glidein Workload Management System (WMS) and the new gLite WMS at CERN to ARC resources. The gLite-ARC interoperability has been improved and the job submission failure rate is about one percent or less. Work on a dedicated ARC plug-in for CRAB continued.

Operations. The ProdAgent ARC plug-in for Monte Carlo production developed by CSC and NDGF has been validated with workflows of several millions of events and is ready for production. The CMS SAM SRM tests were run on the dCache Storage Elements (SEs). NDGF implemented WLCG Site Availability Monitoring (SAM) tests for ARC CEs and these were configured to run on Sepeli and Ametisti and on the dCache SEs. According to the data collected by WLCG the availability and reliability of the Finnish CMS Tier-2 resource in November 2008 was 100 % and 99 % respectively. This excellent result is due to the efficient monitoring and problem resolution of CSC and NDGF. Reporting of WLCG Accounting data was implemented by CSC and NDGF, so that the CMS resource usage is injected to the APEL database. The CMS PhEDEx and Frontier services were run on silo3 in Kumpula by the HIP Technology programme, see the Technology programme text for details. PhEDEx links to all CMS Tier-1 centers were created by the HIP Technology programme and this enabled the Finnish CMS Tier-2 resources to obtained commissioned status. More than 310 TB (Monte Carlo and test) of data was transferred with PhEDEx to Finland. A total of 49 thousand CMS grid jobs using 87 thousand SI2k CPU hours were run in 2008. In addition to this local jobs were also run.