CHIPP Computing Board (CCB) - Minutes of Meeting (Nov 25th 2005) ================================================================ Present: F.Orellana, D.Feichtinger, H.P.Beck, C.Grab, D.Ulmer, R.Bernet, U.Langenegger, C.Haeberli, J. Van Hunen, Xin Wu; Z. Chen (part time). Location: ETHZ - HPK E33; 11:00 - 16:00 1.) Resource Situation at CSCS : (Dominik Ulmer) ---------------------------------------------------- Discussion about the situation arising from Gian Luca Volpato's leaving CSCS. - V. Annoloro will be responsible for the basic system administration of the Phoenix cluster (including system configuration of the basic low level grid services). - Peter Kunszt will take over some of the high level responsibilities (strategic) relating to GRID projects at CSCS, starting in mid. Jan.06. - A new person will be hired by CSCS to take over the system administration, service administration of the cluster and user support. A profile for this person needs to be defined by CSCS and CHIPP. The involved physics experiments/institutes need to discuss their requirements for the position's profile, and then forwarded their input to the CCB-chair (C.Grab). - Intermediate installation of urgent SW-packages: Frederik Orellana will collect a list of experimental SW packages to be installed on the cluster and associated timeframes from the expt's representatives. ---------------------------------------------------------------------- 2) Procurement of new hardware at CSCS: (upgrade of Phoenix) --------------------------------------- - Since the total costs for the new harware will exceed 250000 SFr, an RFP (Request for Proposal) needs to be written. For subsequent acquisition, the option of a "Rahmenvertrag" is aimed for. - The whole process from starting the formulation of the RFP until the new HW can be acquired, may take around ~30-35 weeks (experience of Dominik Ulmer). Therefore it is recommended to start the process soon, before the result of the financial request is known. - A team for producing the RFP needs to be defined; it should be composed by mid Dec 05. Team composition: + one responsible for RFP document + one CSCS SysAd + one CSCS technical expert + one person from CHIPP, collecting all the requirements from the physics experiments' sides (from designated experimental contacts). - Until December 2nd the experimental contacts need to be defined. Name of expt's contact persons to be sent to CG by 2.12. A first round of input from the experimental contacts is requested by end Dec. - For legal reasons there must be just one single contact person for hardware vendor relations. ---------------------------------------------------------------------- 3) Various issues related to GDB/LCG : -------------------------------------- Discussion on VO-Boxes: - Nobody likes the Idea much, but we need to go along with what is needed to carry out the physics tasks. - People running services must be minimal in number and their identities need to be known. - The exact setups and requirements on VO-Boxes are still being composed, so we will need to observe the discussions (the experiment specialists of the CHIPP computing group should pay attention to what requirements their experiment sets forward) - the Grid Acceptable Use Policy: we agree to adhere to it. - The VO Security Policy: we agree to adhere to it. ---------------------------------------------------------------------- 4) Discussion on Tier2 ressource management and scheduling: ----------------------------------------------------------- - Other communities may use the cluster with a lower scheduling priority to prevent wasting of CPU cycles. However it must be guaranteed that the principal users are not inconvenienced. - A document describing the scheduling policy should be prepared which then can be endorsed by the responsibles at CHIPP. - Need to study and understand the technical limitations on scheduling (Batch system technology and Grid user to local user mappings pose difficulties for implementing flexible faireshare policies). ---------------------------------------------------------------------- 5) Status of Tier-3 operation at home institutes: ------------------------------------------------- * University of Bern (C. Haeberli, ATLAS): .......................................... - New computer room containing the two clusters (LHEP and Ubelix). Common setup. Ubelix (some 230 nodes) running NorduGrid on top of Sun GridEngine - Test cluster of dual core opterons (12 CPUs). - new 10 TB RAID system on its way - ATLAS SW: C. Haeberli had to use about 50% of his time for supporting a few users during the production phase in summer. - ATLAS SW: 64bit compiled version slowly evolving. - no LCG software used because maintenance effort prohibitive. * University of Geneva (F. Orellana, ATLAS): ............................................ - will do a massive test production to test the system. - no LCG software used because maintenance effort prohibitive. * University of Zurich (R. Bernet, LHCb): ......................................... - no new HW - no LCG software used because maintenance effort prohibitive. Dirac installed on all machines, and used for MC production; no local analysis. - currently not in need of a dedicated LHCb manager for their SW at CSCS. * ETH, PSI, UZH in CMS (U. Langenegger, CMS): ............................................. - Working at CSCS difficult due to missing CMS specific services for data management. D. Feichtinger has now finished with his responsibilities in the ARDA/EGEE project at CERN and will install and configure the CMS needed services (he will be situated at PSI). - Asgard Cluster being set up with Scientific Linux. - ETH and PSI looking into possibility of having a common Tier3 structure. EPFL (J. Van Hunen, LHCb): .......................... - now ~ 10 FTE, will ramp up to 15 in 2008. - using their cluster privately now, but thinking of converting it to regular Tier3 later on. ---------------------------------------------------------------------- 6.) AOB: -------- - Need to resolve the questions on how to obtain certificates for people and machines. Need to follow up current situation with SWITCH. F. Orellana will take over this responsibility from C. Haeberli. - A status page for the cluster should be hosted by CSCS. (move edginstall to CSCS?) - Christian Haeberli is leaving the Univerity of Bern. The computing board thanks Christian for his engagement and his substantial contributions to our Swiss Tier-2. The new deputy chair of the computing board will be F. Orellana. ---------------------------------------------------------------------- 28.11.2006; D.Feichtinger + C.Grab ----------------------------------