T2_CH_CSCS CMS Information

User Information (mostly OBSOLETE)

• User Information: Help and examples on how to work on the CSCS Tier-2
• Tier-2 to physics groups associations
  • Currently, T2_CH_CSCS is associated to the b-physics group, only

Administrator information for running the local site

• new CMS box (updated 2021)

mostly OBSOLETE INFO:

• CMS Site Log for T2_CH_CSCS
• Data sets to keep/delete table - used in querying users as to what can be deleted
• CMS Software area (structure, setup, verification)
• CMS SiteDB: CMS database with site information. Here we also need to put our resource pledges for each quarter.
  • Q3/2010: CPU 1200 kSI2k (12000 HEPSPEC06 /4 = 3000 kSI2k. From this value 40%), 300 job slots, and 250 TB disk from our 300 TB total share.
  • Q2/2011: CPU 1350 kSI2k (40% of 13488 HEPSPEC06, then using 4 HS06/kSI2k conversion), 520 job slots (556 would have been exact 40% share), and 550 TB disk
• Obsolete CMS site contact's shifter information: Daily interactive checks for our site's health

External admin relevant information

• 2012: Data manager duties
  • Also contains a list of the storage spaces (+size) for a Tier-2
• CMS Computing Operations meeting Wiki / indico page
• Operational Procedures for CMS Sites from the central CMS wiki
• Comp Ops Tier2 guide from the central wiki
• T2 admin guide from the central wiki still ok

CMS Requirements

• CMS has now put some of the most important information on public wiki pages
  • VOMS roles to be mapped by T2 centers
  • limits (ulimits) for worker node resource protection
  • Policies on /user/temp not yet settled
• CREAM-CE (http://grid.pd.infn.it/cream/field.php?n=Main.ReleaseNotes)
  • 2011-02-18 Comp Meeting: glexec + ARGUS should be operational by 2011-06-30.
  • 2011-02-21 I now added cream02.lcg.cscs.ch to our CMS sitedb description
  • CMS recommends to start deploying gLite 3.2 (SL5) CREAM 1.6.3 now (2011-01) and in any case before March 31st (FacOps Meeting 2011-01-17)
  • CMS suggests to keep the LCG-CE until SAM implications, etc. better understood
  • CMS site tests will become critical by end of Jan (i.e. if a site has a CREAM-CE, it must work by then). A site may run both LCG and CREAM CEs, but each one must work then. It is ok to run just one of the services.
• Batch system related requirements
  • Heavy ion production users
    • CMS wants to introduce role=hiproduction for the MC production of heavy ion events (q.v. Talk by T. Kress). We just can map them to the existing cmsprod pool accounts (depends on local communities. TODO: ask Swiss community).
  • Priority users
    • A limited number of users for every CMS physics group can take the Priority User VOMS role (role=priorityuser). These should be mapped to a group of local system users (e.g. priorityuser01, ...). These should get a higher priority than normal CMS users. Questions can be asked from hn-cms-analysisoperations@cern.ch.
    • OBSOLTETE: central Twiki page indicating readiness, to be filled out by us. Also, look at talk by T. Kress at FacOps 2009-09-28
  • Pilot users: Normal jobs using the pilot framework
• CMS job I/O requirements
  • 2009-06-18 Ian Fisk reports on CMSSW I/O campaign in PADA meeting. Aim is to increase CPU efficiency while trying to keep I/O as low as possible.
  • 2009-03-30 Ian Fisk's Talk on specific dCache issues mentions CMS I/O bound jobs requiring 3 MB/s of data.
• Storage related requirements
  • Storage organization and sizing for CMS
    • Note: on 2011-05-20 confirmed to A. Fanfani that b-phys group has now 100 TB for data set storage
  • requirements on the namespace setup (smaller table)
  • VOMS role=production must be able to write to /store/results/ path
  • VOMS role=priorityuser must be able to write into /store/group/ path

Interesting Talks and links on Tier-2 relevant technologies

• Tier-2 Storage organization
  • Ian Fisk's slides from Computing Integration Meeting Jul 16th 2008. This defines a schema of storage classes and sizes.
  • Ian Fisk's slides from PADA Meeting of Sep 4th 2008 with Tier-2 to physics groups associations.
  • Ian Fisk's slides from CMS week of Sep 25th 2008. Contain descriptions of storage policies and group mappings.
• Three tech talks on SE technologies from the CMS offline workshop 2009-04:
  • GPFS/Storm
  • Lustre/dCache
  • Hadoop
    • Hadoop OSG site: https://twiki.grid.iu.edu/bin/view/Storage/Hadoop
• dcache to Lustre/Storm comparison by Giacinto Donvito at the CMS offline workshop 2009-10
• Information on pilot jobs
• CVMFS: HTTP/Fuse based file system for deploying software
  • Paper by Jakob Blomer and Predrag Buncic
  • CERNVM project page

old information

• Status of Summer08 Monte Carlo production
• obsolote Analysis Group page on CERN twiki

-- DerekFeichtinger - 24 Apr 2009