Presentation is loading. Please wait.

Presentation is loading. Please wait.

INFN Tier1/Tier2 Cloud WorkshopCNAF, 22 November 2006 Conditions Database Services How to implement the local replicas at Tier1 and Tier2 sites Andrea.

Similar presentations


Presentation on theme: "INFN Tier1/Tier2 Cloud WorkshopCNAF, 22 November 2006 Conditions Database Services How to implement the local replicas at Tier1 and Tier2 sites Andrea."— Presentation transcript:

1 INFN Tier1/Tier2 Cloud WorkshopCNAF, 22 November 2006 Conditions Database Services How to implement the local replicas at Tier1 and Tier2 sites Andrea Valassi (CERN IT-PSS)

2 2 22 2 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 Conditions DB in the 4 experiments ALICE –AliROOT (Alice-specific software) for time/version handling –ROOT files with Alien file catalog CMS –CMSSW (CMS-specific software) for time/version handling –Oracle (via POOL-ORA C++ API) with Frontier web cache ATLAS and LHCb –COOL (LCG AA common software) for time/version handling –Oracle, MySQL, SQLite, Frontier (via COOL C++ API)

3 3 33 3 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 ALICE – use Alien/ROOT (no DBs) ALICE has no special needs for database deployment and replication at Tier1 and Tier2 sites for managing its conditions data

4 4 44 4 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 CMS – use Oracle/Frontier Oracle+Tomcat at Tier0 and Squid caches at Tier1 and Tier2 already set up to access CMS conditions data during CSA06 NB Cache consistency control - caches must be refreshed periodically (querying T0 db)

5 5 55 5 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 Atlas and LHCb – use COOL COOL: LCG Conditions database –Common development of IT-PSS, LHCb and Atlas –Handle time-variation and versioning of conditions data Four supported relational technologies (via CORAL) –Oracle and MySQL database servers –SQLite files –Frontier (read-only ): Squid + Tomcat + Oracle server COOL service deployment model –Based on generic 3D distributed db deployment model Oracle at Tier0 and Tier1 (with distribution via Oracle Streams) Other technologies elsewhere if at all needed –Details depend on each experiment’s computing model

6 6 66 6 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 LHCb computing model COOL only stores the conditions data needed for event reconstruction –Oracle at Tier0 –Oracle at Tier1’s (six sites) –Streams replication T0-T1 –COOL not needed at Tier2’s (only MC production there) –SQLite files may in any case be used for any special needs (Marco Clemencic, 3D workshop 13 Sep 2006)

7 7 77 7 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 LHCb – COOL service model Two servers at CERN – essentially for online and offline –Replication to Tier1’s from the online database is a two-step replication –Server at the pit managed by LHCb; offline server managed by IT-PSS FZK RAL IN2P3 CNAF SARA PIC (Marco Clemencic, 3D workshop 13 Sep 2006) COOL (Oracle)

8 8 88 8 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 LHCb – status and plans Oracle and Streams replication set up and tests –Oracle servers at PIT (single-instance prototype), CERN offline (test RAC) and three Tier1 sites (FZK/Gridka, RAL, IN2P3) –Tested two-step Streams replication between PIT-CERN-Tier1 –Tested replication throughputs much higher than expectations All OK with 100 IOVs/sec for one hour and 1 IOVs/sec for 24 hours –In progress: stress-test T1 read access; test tag replication latency Future milestones –By March ‘07: add CNAF (Dec), Nikhef/SARA (Jan), PIC (Mar) –By March ’07: integration with CORAL LFC replica svc and latest COOL DB lookup and secure user authentication using grid certificate (also needs secure SSH data transfer in LFC – deployment expected Jan-Feb ‘07) –April ’07: production using CERN offline and T1s (pit still in test mode)

9 9 99 9 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 Atlas – COOL service model COOL Oracle services at Tier0 and ten Tier1’s –Two COOL servers at CERN for online/offline (similar to LHCb) Online database within the Atlas pit network, but physically in the CC –In addition: Oracle (no COOL) at three ‘muon calibration center’ Tier2’s Online OracleDB Offline master CondDB Tier-0 recon replication Tier-1 replica Tier-1 replica Online / PVSS / HLT farm Tier-0 farm Dedicated 10Gbit link ATLAS pitComputer centreOutside world CERN public network Calibration updates Streams replication ATLAS pit network (ATCN) gateway (Sasha Vaniachine and Richard Hawkings, 3D workshop 14 Sep 2006) FZK ASGC BNL RAL IN2P3 CNAF TRIUMF Nordugrid PIC SARA

10 10 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 Atlas – muon calibration centers (Sasha Vaniachine and Joe Rothberg, 3D workshop 14 Sep 2006) Muon calibration centers: Roma, Michigan, (Munich) Streams set up between Michigan (source) and CERN (target)

11 11 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 Atlas – status and plans Oracle and Streams replication set up and tests –Oracle servers at CERN online and offline (test RACs) and five Phase1 Tier1 sites (FZK/Gridka, BNL, ASGC/Taiwan, IN2P3, RAL) - plus Triumf TCP and protocol tuning for BNL and ASGC: 400 -> 3200 LCR/sec –Tested two-step Streams replication between online-offline-Tier1 Future milestones –By December ‘06: add CNAF –February ’07: “CDC” production with all six Tier1’s –Three “Phase 2” Tier2 sites to join later (Nordugrid, SARA, PIC) Open issues for Atlas –Can the achievable replication throughput meet the Atlas requirements? Intrinsic Streams limitation: single row updates in the apply step Need detailed throughput requirements – take as much as Streams can do? Considering transportable tablespaces for TAGS (not possible for COOL?) –Replication to Tier2’s COOL ‘dynamic replication’, e.g. to MySQL – under development Evaluating COOL Frontier backend (performance, cache consistency…)

12 12 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 Streams downstream capture Objective –Source database (CERN) isolation against network or database problems at the replicas Status and plans –Blocking Oracle bug solved two weeks ago –Four nodes allocated –Production setup in December with two servers (Atlas/LHCb) and two spares Eva Dafonte Perez (IT-PSS)

13 13 Andrea Valassi Conditions Data ReplicationCNAF, 22 November 2006 Summary ALICE: use ROOT and Alien (no db services outside CERN) CMS: Oracle and Frontier at T0, Squid web caches at T1/2 –Already set up and successfully tested during CSA06 –No database service is required outside CERN for CMS LHCb (COOL): Oracle at T0 and T1 with Streams replication –Two servers at CERN: one at the PIT (online) and one in the CC (offline) –No database service is required at T2’s for LHCb –Production with all 6 T1 sites in April 2007 Atlas (COOL): Oracle at T0 and T1 with Streams replication –Two servers at CERN, both in the CC (online server in the online network) –Production with 6 ‘Phase1’ T1 sites in February 2007 –Potential problem with Streams throughput – need detailed requirements –No decision yet about Tier2 deployment – evaluating MySQL and Frontier


Download ppt "INFN Tier1/Tier2 Cloud WorkshopCNAF, 22 November 2006 Conditions Database Services How to implement the local replicas at Tier1 and Tier2 sites Andrea."

Similar presentations


Ads by Google