300 likes | 482 Views
C. Zampolli ALICE Offline Week, 25-27 March 2013. Calibration. Ever tried. Ever failed. No matter. Try Again. Fail again. Fail better. (S. Beckett). ALICE current Calibration(s ). Data. DAQ DA. Shuttle. OCDB. Data. CPass0. OCDB. Data. CPass1. OCDB. Manual Calibration. Data.
E N D
C. Zampolli ALICE Offline Week, 25-27 March 2013 Calibration C. Zampolli - Calibration - Offline Week
Ever tried. Ever failed. No matter. Try Again. Fail again. Fail better. (S. Beckett)
ALICE current Calibration(s) Data DAQ DA Shuttle OCDB Data CPass0 OCDB Data CPass1 OCDB Manual Calibration Data Reconstruction ESDs for analysis C. Zampolli - Calibration - Offline Week
Calibration strategy – CPass0 + CPass1 CPass0 CPass1 Chunk 0 Chunk N Chunk 1 Chunk 0 Chunk N Chunk 1 … … Alien job Alien job Alien job Alien job Alien job Alien job Alien job Alien job Used for 2012 and 2013 data taking snapshot OCDB snapshot reconstruction reconstruction ESDs ESDs Calib. train QA train Calib. train Merging + OCDB entires Merging + OCDB entires C. Zampolli - Calibration - Offline Week OCDB Update OCDB Update
2012 Calibration Scheme Calibration period, Δ cPass0 cPass1 Daily QA from detectors Savannah tickets and communication between QA coordination and detectors Used for 2012 and 2013 data taking (some ad-hoc modifications for 2013 pA data) QA weekly meeting QA weekly meeting +7d ~3 weeks Only Savannah tickets discussed, to be then handled by offline + calibration groups Manual update (1w) Max delay between end of data taking and validation of calibration ~3 weeks vPass QA on vPass Main QA C. Zampolli - Calibration - Offline Week
2012-2013 summary pA periods Completed Ongoing Not foreseen Very soon Delayed to give priority to pA Soon C. Zampolli - Calibration - Offline Week
When and why the timeline was not followed • Resubmissions of (any) failed “parts” • Many reasons, e.g.: • Trigger aliases problem • Failures of calibration for some detector which prevented CPass1 from starting • Much better situation since we use the return code • Negative: warning, calibration not ok, but expected, so please continue • Positive: ERROR, calibration not ok, not expected, DO NOT continue • 0: Everything ok • Some grid issues, e.g.: • Disks full • Problems with one site • OCDB problem • CPass0 stopped (e.g. in June) to fix code, synchronize it C. Zampolli - Calibration - Offline Week
When and why the timeline was not followed • Resubmission of an entire period • LHC12d (2x!), LHC12h, LHC12i • To account for detectors code bugs, unexpected behavior of calibration with new data taking conditions • Too short period • Need to be merged with previous or subsequent ones • Lack of feedback from detectors • But people need holiday… • …and much better since Autumn 2012 A huge thanks to all the people who have worked (hard) over the last month (and year) to “turn” the ALICE data into the input for Physics analyses! C. Zampolli - Calibration - Offline Week
When and why the timeline was not followed • Resubmission of an entire period • LHC12d (2x!), LHC12h, LHC12i • To account for detectors code bugs, unexpected behavior of calibration with new data taking conditions • Too short period • Need to be merged with previous or subsequent ones • Detectors needing more than 1 week to assess calibration readiness • And in pA… unexpected problems… see later (unfortunately, not enough space in one slide ) C. Zampolli - Calibration - Offline Week
Issues during pA • Long runs: in principle not an issue, but… • Online calibration • Reconstruction time • Merging time • fixed very quickly, TPC limiting number of chunks to be merged for them • SPD hs misaligned in time at the end of run • only good chunks reconstructed • Low reconstruction efficiency • partially fixed in v5-03-Rev-17 • TPC online calibration in rare periods • recovered at CPass1, or need to re-run • HMPID not reconstructed in CPass1 for LHC13d, e, f • fixed asap, but already in LHC13f • TPC HV trips, HV voltage changed during the run (30/140 runs) in LHC13d, e, f • TPC code modified to take this into account in v5-03-Rev-15 • Not sure how to treat this in MC… C. Zampolli - Calibration - Offline Week
Issues during pA – II • Analysis: a source of precious information, which allowed to spot problems • First event in chunk not calibrated for TPC • Related to the use of the HLT seeds, introduced since 2013 processing • Fixed in v5-03-Rev-15 • Chunks without any TPC information • Related to the way HLT used OCDB, and specifically to when an entry was not there • Fixed in v5-03-Rev-15 C. Zampolli - Calibration - Offline Week
Essentialities • A lot of synergies within different “communities” • Offline • Reconstruction • GRID (including job (re)submission) • MonALISA • Detector experts • Tuning of calibration (especially TPC) • Fixes of code • Updates of code • But not only offline… (see TPC during LHC13d,e) • QA • Fundamental to spot problems • Run Coordination and Trigger Coordination • Constant communication C. Zampolli - Calibration - Offline Week
Flaws in the procedure? • Absolutely not! But if we want to talk about room for improvement… • Reduce manual interventions during the processing • More stable calibrations • Calibrations more robust wrt changes in the running conditions • More stable code • Check 3x before committing code, test it 4x before using it in production • Reduce the time between data taking and calibrated data (i.e. Physics Passes) • Increase fraction of online calibrations • Reduce inter-dependencies between detectors • Automatize QA as much as possible C. Zampolli - Calibration - Offline Week
Up to now… Ever tried. Ever failed. No matter. Try Again. Fail again. Fail better. (S. Beckett) Nothing ventured, nothing gained. Its’ a tough job, but… Really? C. Zampolli - Calibration - Offline Week
Something is “easier” than… • Reprocessing • Using the same schema as up to now • CPass0 + CPass1 + VPass + Ppass • Cannot (obviously) rely on Trigger Aliases • Some pre-requisites are necessary • Memory consumption Fixed! C. Zampolli - Calibration - Offline Week
Something is “easier” than… • Reprocessing • Using the same schema as up to now • CPass0 + CPass1 + VPass + Ppass • Cannot (obviously) rely on Trigger Aliases • Some pre-requisites are necessary • Memory consumption Fixed! • Some issues with calibration Fixed! • Detector simulation should match Data • Could analysis be used to assess if something is missing? • Review of Savannah open bugs… C. Zampolli - Calibration - Offline Week
…something else Pant! Run3… Pant! Run2… C. Zampolli - Calibration - Offline Week
Preparation for Run2/Run3 See Document
Main goal of Upgrade C. Zampolli - Calibration - Offline Week
Run2 • Detectors should try to move as much as possible online • Detector algorithms • Due to the dependency on tracking (TPC), everyone should rely on an online TPC calibration (within HLT) and an online reconstruction • The Shuttle will then continue to export the Condition Data to the OCDB • After the run, possibility to run (one single!) CPass, and perform manual calibration • Then (likely) VPass, Ppass • …but the data before CPass should be good enough for a first physics Manual Calibration Data Reconstruction ESDs for analysis Data Online Processing Shuttle OCDB Data CPass0 OCDB C. Zampolli - Calibration - Offline Week
Run3 • No more CPass-espossible • Only second-order corrections possible • Use O2 farm to perform as much online calibration as possible, and online reconstruction • As for Run2, due to the dependency on tracking (TPC), everyone should rely on an online TPC calibration (within HLT) and an online reconstruction C. Zampolli - Calibration - Offline Week
Run3 • No more CPass-espossible • Only second-order corrections possible • Use O2 farm to perform as much online calibration as possible, and online reconstruction • As for Run2, due to the dependency on tracking (TPC), everyone should rely on an online TPC calibration (within HLT) and an online reconstruction Online Processing : Calibration and Reconstruction Data ESDs for analysis OCDB C. Zampolli - Calibration - Offline Week
The TPC proposal See talk this afternoon J. Theader at O2 meeting C. Zampolli - Calibration - Offline Week
The TPC proposal - consideration See talk this afternoon Not the standard OCDB, but something available online; standard OCDB updated at the end of the run/fill J. Theader at O2 meeting C. Zampolli - Calibration - Offline Week
The TPC proposal - continued See talk this afternoon • Calibration performed in “Calibration intervals” • TPC would extrapolate the calibration obtained in the previous calibration interval • Stability in time to be able to extrapolate • First events (not calibrated) stored anyway at Run2, flagged somehow at Run3 • Data would be TPC-calibrated when reconstructed • All the other detectors may use TPC-calibrated data for their own calibration C. Zampolli - Calibration - Offline Week
Implications for detectors • Mainly detectors participating in CPass are affected • TPC will move online anyway • TRD • TOF • T0 • MeanVertex • SDD C. Zampolli - Calibration - Offline Week
Conclusions • Procedure used for calibrating 2012-2013 data (CPass0 CPass1 VPassPPass) subject to many manual interventions… • Sensitive to many parameters • …but in general quite successful • To be used for re-processing for old data • Preparation for Run2/Run3 relies on online calibrations • Basic ingredient: TPC – who already started planning for this C. Zampolli - Calibration - Offline Week