Data processing Offline review Feb 2, 2011. 2 Productions, tools and results Three basic types of processing RAW MC Trains/AODs I will go through these.

Slides:



Advertisements
Similar presentations
5/2/  Online  Offline 5/2/20072  Online  Raw data : within the DAQ monitoring framework  Reconstructed data : with the HLT monitoring framework.
Advertisements

– Unfortunately, this problems is not yet fully under control – No enough information from monitoring that would allow us to correlate poor performing.
DATA PRESERVATION IN ALICE FEDERICO CARMINATI. MOTIVATION ALICE is a 150 M CHF investment by a large scientific community The ALICE data is unique and.
Trains status&tests M. Gheata. Train types run centrally FILTERING – Default trains for p-p and Pb-Pb, data and MC (4) Special configuration need to be.
ALICE Operations short summary and directions in 2012 Grid Deployment Board March 21, 2011.
ALICE Operations short summary LHCC Referees meeting June 12, 2012.
ALICE Operations short summary and directions in 2012 WLCG workshop May 19-20, 2012.
Desktop Security: Worms and Viruses Brian Arkills, C&C NDC-Sysmgt.
Chiara Zampolli in collaboration with C. Cheshkov, A. Dainese ALICE Offline Week Feb 2009C. Zampolli 1.
MC, REPROCESSING, TRAINS EXPERIENCE FROM DATA PROCESSING.
L3 Filtering: status and plans D  Computing Review Meeting: 9 th May 2002 Terry Wyatt, on behalf of the L3 Algorithms group. For more details of current.
Status of the production and news about Nagios ALICE TF Meeting 22/07/2010.
Costin Grigoras ALICE Offline. In the period of steady LHC operation, The Grid usage is constant and high and, as foreseen, is used for massive RAW and.
Production status Preparation for HI running ALICE TF+AF November 14, 2010.
Analysis infrastructure/framework A collection of questions, observations, suggestions concerning analysis infrastructure and framework Compiled by Marco.
Offline report – 7TeV data taking period (Mar.30 – Apr.6) ALICE SRC April 6, 2010.
DQM status report Y. Foka (GSI) Offline week from pp to PbPb.
Infrastructure for QA and automatic trending F. Bellini, M. Germain ALICE Offline Week, 19 th November 2014.
PWG3 Analysis: status, experience, requests Andrea Dainese on behalf of PWG3 ALICE Offline Week, CERN, Andrea Dainese 1.
Andrei Gheata, Mihaela Gheata, Andreas Morsch ALICE offline week, 5-9 July 2010.
Analysis trains – Status & experience from operation Mihaela Gheata.
5/2/  Online  Offline 5/2/20072  Online  Raw data : within the DAQ monitoring framework  Reconstructed data : with the HLT monitoring framework.
CERN – Alice Offline – Thu, 20 Mar 2008 – Marco MEONI - 1 Status of Cosmic Reconstruction Offline weekly meeting.
T0 offline status Alla Maevskaya for T0 team 8 March 2011 ALICE offline week.
Moritz Backes, Clemencia Mora-Herrera Département de Physique Nucléaire et Corpusculaire, Université de Genève ATLAS Reconstruction Meeting 8 June 2010.
Resource Request Tracking (RRT) The first component of Wildfire One June 2015.
1 Checks on SDD Data Piergiorgio Cerello, Francesco Prino, Melinda Siciliano.
A. Gheata, ALICE offline week March 09 Status of the analysis framework.
AliRoot survey: Analysis P.Hristov 11/06/2013. Are you involved in analysis activities?(85.1% Yes, 14.9% No) 2 Involved since 4.5±2.4 years Dedicated.
Gustavo Conesa ALICE offline week Gamma and Jet correlations analysis framework Short description, Status, HOW TO use and TO DO list 1/9.
Predrag Buncic CERN Future of the Offline. Data Preparation Group.
M. Gheata ALICE offline week, October Current train wagons GroupAOD producersWork on ESD input Work on AOD input PWG PWG31 (vertexing)2 (+
PWG3 Analysis: status, experience, requests Andrea Dainese on behalf of PWG3 ALICE Offline Week, CERN, Andrea Dainese 1.
PWG3 analysis (barrel)
Physics selection: online changes & QA M Floris, JF Grosse-Oetringhaus Weekly offline meeting 30/01/
CWG7 (reconstruction) R.Shahoyan, 12/06/ Case of single row Rolling Shutter  N rows of sensor read out sequentially, single row is read in time.
Analysis experience at GSIAF Marian Ivanov. HEP data analysis ● Typical HEP data analysis (physic analysis, calibration, alignment) and any statistical.
Analysis train M.Gheata ALICE offline week, 17 March '09.
M. Gheata ALICE offline week, 24 June  A new analysis train macro was designed for production  /ANALYSIS/macros/AnalysisTrainNew.C /ANALYSIS/macros/AnalysisTrainNew.C.
V5-01-Release & v5-02-Release Peter Hristov 20/02/2012.
Offline Weekly Meeting, 24th April 2009 C. Cheshkov & C. Zampolli.
RAW and MC production cycles ALICE Offline meeting June 21, 2010.
 offline code: changes/updates, open items, readiness  1 st data taking plans and readiness.
LHCb Computing activities Philippe Charpentier CERN – LHCb On behalf of the LHCb Computing Group.
Some topics for discussion 31/03/2016 P. Hristov 1.
D. Elia (INFN Bari)Offline week / CERN Status of the SPD Offline Domenico Elia (INFN Bari) Overview:  Response simulation (timing info, dead/noisy.
AliRoot survey: Calibration P.Hristov 11/06/2013.
LEGO train limits Offline week, 31/03/2016 LB 1. Incident of 22/03/2016 A train operator has submitted 107 LEGO trains in one go – These resulted in more.
V4-19-Release P. Hristov 11/10/ Not ready (27/09/10) #73618 Problems in the minimum bias PbPb MC production at 2.76 TeV #72642 EMCAL: Modifications.
DAQ thoughts about upgrade 11/07/2012
ANALYSIS TRAIN ON THE GRID Mihaela Gheata. AOD production train ◦ AOD production will be organized in a ‘train’ of tasks ◦ To maximize efficiency of full.
CALIBRATION: PREPARATION FOR RUN2 ALICE Offline Week, 25 June 2014 C. Zampolli.
1 14th June 2012 CPass0/CPass1 status and development.
V4-18-Release P. Hristov 21/06/2010.
Jan Fiete Grosse-Oetringhaus
Calibration: preparation for pa
Analysis trains – Status & experience from operation
DPG Activities DPG Session, ALICE Monthly Mini Week
Production status – christmas processing
ALICE analysis preservation
fields of possible improvement
v4-18-Release: really the last revision!
New calibration strategy – follow-up from the production review
Offline shifter training tutorial
Analysis Trains - Reloaded
Disk capacities in 2017 and 2018 ALICE Offline week 12/11/2017.
Analysis framework - status
ALICE Computing Model in Run3
TPC status - Offline Q&A
Offline framework for conditions data
Presentation transcript:

Data processing Offline review Feb 2, 2011

2 Productions, tools and results Three basic types of processing RAW MC Trains/AODs I will go through these with as few details as possible I will not discuss the AliRoot evolution This deserves a long separate presentation

3 MC productions MC productions (on behalf of PWGs) Request in AliRoot Savannah (as tasks) Macros and JDLs (supplied by the requestor) in MonALISA The priority and validity of the request are discussed by PB Exceptions – for short and urgent productions this step is usually short-circuited, e.g. first physics, overnight, weekend (the Friday evening syndrome)

4 MC production statistics MC productions (on behalf of PWGs) All normal requests and exceptions have been handled We have run 161 different MC production cycles throughout 2010 (effectively one each 2 days), quite unevenly spread Total of 969Mio events

5 MC production in 2010 Recent PB decisions Reduction of MB cycle to 10% of RAW statistics Compensated by PWG specific signals/configuration requests The effective reduction in processing time/resources in general will be negligible, if any Likely, the number of MC production requests will grow this year…

6 MC production – what follows QA PWG1 train AODs filtering train Luckily all of the above without tenders (yet) ESD and AOD are still consistent The process is simpler than for RAW data, but still not fully automatic – AOD trains always have ‘last minute changes’ and usually are delayed with respect to the ESD processing The usual culprit is excessive memory consumption

7 RAW data productions Two types – Physics and Special Pass 1 (with or without Pass0) – automatic, quasi online, always latest AliRoot revision Pass2+ and special – AliRoot savannah as tasks The tasks usually follow the status of calibration and code updates necessary for the pass to start A complicated, long and ever changing procedure…

8 RAW data productions (2) Special passes Detector calibration runs Usually no problems with these Specific ‘quality check’ physics runs or detector calibration passes (calorimetry) With updated AliRoot version and/or OCDB Either precursors of or post-Pass(x) Tend to be repetitive and follow each other quickly i.e. ‘Try again after a bug fix or OCDB update’ Not necessarily small statistics

9 RAW data passes and statistics Total of 52 RAW data passes in ,197,284,148 events (as of last evening) Does not include the Pass0 processing, but includes the detector calibration data

10 RAW data reco specifics More complex than MC, 3 basic reasons Detector code is better ‘adapted’ to MC processing More events per job – memory leaks and reconstruction code exceptions are appearing here The run conditions change during the period, the code needs to be adapted (more on this on the next slide)

11 RAW data reco specifics (2) Pass 1 processing Run always with the latest available AliRoot revision This catches running reco problems, as seen during the quasi-online reconstruction Fixes follow weekly Ideally (and this is also in the CM) Pass2 should follow shortly after Pass1, effectively the latest code, only calibration update Same AliRoot revision, should be OK…

12 RAW data reco specifics (3) What happens in reality Long period of calibration after the period is over Numerous changes in the code, usually relevant to the next running period And often specific to it OCDB updates – same as above As a consequence for Pass2 the conditions have changed considerably Case in point LHC10e(2)

13 RAW data – what happens next Analysis trains PWG1 QA train – running after each pass Tag merging AOD filtering (several trains), delayed with respect to production up to a month These are with tenders… Running with AN tag And are repeated several times

14 Trains statistics 280 Analysis trains in 2010 about 1/2 are merging trains (reduces fragmentation) Each train has a unique configuration – detectors included/excluded depending on readiness Always on the bleeding edge of AliRoot AN tag… None of the trains are calibration…with the exception of Pass0

15 Summary of production in MC cycles 52 RAW cycles 280 analysis trains Almost each of the above has unique configuration We’ve managed reduce the macros complexity only for Pass1(+) reco A small demo of the cascading of tasksdemo

16 General risk factors Configuration Error in the processing macros OCDB Wrong or out-of-synch update Code Insufficient testing prior to submission All of the above – never enough time New hooks and pre/post-processing are being added continuously

17 General risk factors (2) Validation and tests - the results are not verified quickly enough Given the vast amount of output, this is not surprising The ‘validation pause’, introduced long ago is simply impractical The production must finish for the next step (or cycle) to begin

18 Beware of ‘good ideas’ Accidental deletion of 900 GeV data Someone in 2007 had the ‘good idea’ to introduce recycle tape pools Basically, automatic data deletion in a storage system was entrusted to a script It was just a question of time before something important went there… and it did

19 Summary The ALICE production system was able to cope with all processing requests in 2010 Which were ready to start We have a relatively flexible approach and tools to carry out the production This includes the code/OCDB update practices But this seem to just increase the complexity of the tasks

20 Summary (2) The general trend is to push processing toward later tasks Online calibration –> Pass0 Reco QA -> PWG1 train AOD filtering -> AOD filtering (with tender) User analysis tasks -> ESDs and AODs with tender Consequences – see Federico’s and Yves’ slides