Introduction L. Poggioli, LAL • ATLAS latest news • Actions list Inputs from recent meetings – LCG-TECH ftf 16/04 https://indico.in2p3.fr/conferenceDisplay.py?confId=9731 – Pre-GDB 13/05 http://indico.cern.ch/event/272787/ CAF_T2_150514 Luc 1
Pledges: C-RSG feedback (1) CPU reduction: HLT usage should Be pledged CAF_T2_150514 Luc 2
Pledges: C-RSG feedback (2) Potential budget crisis? • Run-2 LHC parameters likely to be pessimistic • Flat budget model limitations (C-RSG study ongoing) CAF_T2_150514 Luc 3
Activities since last CAF OK but big fluctuations from production CAF_T1_150514 Luc 4
Production last month (1) SW not ready / Lack of jobs to process CAF_T1_150514 Luc 5
Production forecast Claire, Wolfgang ADC weekly 29/04 • Based on delivery of new sw release 19.x.x end May • Till then NO MCORE activities CAF_T1_150514 Luc 6
MCORE • Today After clarification with Andrej, Simone, Andreu – MC12 simul & reco 1core – MC14 xcore – Xcore: 25-30% total • Soon – Only MC14 – Xcore: 60-70% total • “Big” sites – Asked to deploy xcore dynamically (a priori easy in Torque) – Timescale: 1-2 months (i.e. to be ready for DC14) CAF_T2_150514 Luc 7
JIRA, RUCIO • Savannah -> JIRA migration – Almost done (DPD project remains to move) • RUCIO – Migrated clouds: ALL but US, CERN, NDGF – Full commissioning stress test (on real but not official data) 20/05-End June – All DDM endpoints have new naming except /SAM/ subdirectory • Files and directories not following this convention are out of DDM catalogs (dark data) and can be removed • To be done per site/squad CAF_T2_240314 Luc 8
perfSonar • All FR-sites have deployed perfSonar • Monitoring http://maddash.aglt2.org/maddash-webui/ Bandwidth Latency • More a site tool than a VO tool • F. Schaer is following – eg Firewall issues, asymmetries, inconsistencies CAF_T2_240314 Luc 9
XrootD/FAX/HTTP (1) • ATLAS priorities to sites (for T1s/T2Ds) 1) Enable xrootD data access 2)Enable FAX 3) Enable HTTP/WebDaV data access • Done for most FR-T2Ds& T1 SSB dashboard Rob, pre-GDB – LPC ongoing, LPSC? • FAX: 48 sites in – Failover mode • 241 queues (prod & ana) • Tiny %network used – Overflow mode • Leave data, move job • Tested for US sites CAF_T2_240314 Luc 10
XrootD/FAX/HTTP (2) • WebDAV Cédric, pre-GDB – All functionalities to manipulate data • RUCIO aware of sites DT • RUCIO knows where to find replicas – Supports FTS (candidate to replace SRM) • When final RUCIO migration done – Possible to access files via WebDaV – 62 sites today, 329 endpoints – Access via RUCIO redirector or Metalinks • Efforts needed – SAM tests, Monitoring CAF_T2_240314 Luc 11
Remote access to local storage • ATLAS recommendation Campana, Elmheuser, Manoulis – xrootd direct access for analysis – xrdcp for production (copy-to-scratch) • For dpm sites – Most sites are using copy-to-scratch with lcg-cp, but should be encouraged to move to xrootd • More efficient than gridftp copy-to-scratch • RFIO broken for directIO • For dCache – Analysis: Xrootd & dcap similar performance – Prod: xrdcp (eg at CC) LCG-TECH_16042014 Luc Poggioli 12
Actions List (1) • FAX – Follow test jobs results & understand • perfSonar – Follow BW & latency performance • Xrootd DA for analysis queues – If no objection from sites, proceed • MCORE – Try with IRFU, TOKYO – Dynamic • Pledges deployment? CAF_T2_240314 Luc 13
Actions List (2) • ARC CE – Machines in // Try CPPM, LPSC ? • Sites – Analysis queue at Beijing Done – Romania: RO-07 as T2D • Support – DAST situation improved (eg Laurent has joined) – Squad • No news bad news • Recontact all sites (eg Saclay, LAPP, LPC) • Next week: HEPIX 19-23 May at LAPP CAF_T2_240314 Luc 14
Recommend
More recommend
