130 likes | 328 Views
Offline calibration. Offline Weekly Meeting, 24th April 2009 C. Cheshkov & C. Zampolli. Basic Guidelines. Calibration and Alignment should always be AliAnalysisTasks Reuse of the existing Analysis framework All the calibration/align data will be stored in ESD/ESD Friends/AOD files
E N D
Offline calibration Offline Weekly Meeting, 24th April 2009 C. Cheshkov & C. Zampolli
Basic Guidelines • Calibration and Alignment should always be AliAnalysisTasks • Reuse of the existing Analysis framework • All the calibration/align data will be stored in ESD/ESD Friends/AOD files • No additional file • No new Objectinvented C. Cheshkov & C. Zampolli
Calibration Step – 1 Chunk 0 Chunk 1 ... Chunk i Chunk i+1 Chunk i+2 ... Reconstruction (inside the event loop) Calibration/Alignment AliAnalysisTasks Not necessarily AliAnalysisTasks, but useful like this (it can be run on files) ESDFriends i ESD 0 ITS TRD ... Objects that scale with the n. of events! (branches of the tree for ESD friends) C. Cheshkov & C. Zampolli
Calibration Strategy – 1 – comments • Offline Calibration – as well as Alignment – could be organized as “calibration/alignmenttasks” to be run during reconstruction: • Within AliReconstruction, in the event loop • Every detector providing its own task • Every detector wagon should produce Calibration/Alignment data, to be stored in the ESD or ESD Friends (in branches) • If filling ESD friend branhces, only a selected fraction of events will have the calib/align ESD friends branches filled C. Cheshkov & C. Zampolli
Calibration Step – 2 Chunk 0 Chunk 1 ... Chunk i Chunk i+1 Chunk i+2 ... reconstruction ESD i ESD-Friends i Calibration/Alignment train of AliAnalysisTasks ... TRD ITS ESD Friends i ITS TRD ... Objects that do not scale with the n. of events! N.B.: the wagons are the different tasks! C. Cheshkov & C. Zampolli
Calibration Step – 2 – comments • Offline Calibration – as well as Alignment – could be organized as a “calibration/alignmenttrain” to be run immediately after the reconstruction – in the same job, or in the same aliroot session: • Every detector providing its own wagon • Single loop over the events instead of once per detector calibration procedure • Running on the Worker Node on local files ESDFriend available • Every detector wagon should produce Calibration/Alignment data in histogram-like format (not scaling with n. of events), to be stored in the ESD Friends (in folders) C. Cheshkov & C. Zampolli
Getting the Calibration Constants Merging/Analysis procedure to produce a final file ESD 0 ESD ... ESD i ESD i+1 ESD ... AOD 0 AOD... AOD i AOD i+1 AOD ... N.B.: the wagons are the calibration objects written (in the ESD friends) before! ESD Friends i+1 ESD Friends... ESD Friends... ESD Friends 0 ESD Friends i automatically ITS ITS ITS ITS TRD TRD TRD TRD ... ... ... ... ITS TRD ... ...i... ... 0... The expert... OCDB Updating C. Cheshkov & C. Zampolli
Getting the Calibration Constants – comments • Merging the information coming from the different runs • The procedure is a train, with the wagons being the various ESD/ ESD friends/ AOD files from the various runs • Calculation of the calibration constants according to the result of the merging procedure either manual or automatized C. Cheshkov & C. Zampolli
Some Details • ESD friends should be modified in order to fit the new schema • Become independent from ESDs friend tree • Reduction in size • Every detector should evaluate the number and type of events needed and the size of data to be stored • ESD friends file should be always kept, but with ESD friend objects saved only for the requested fraction of event C. Cheshkov & C. Zampolli
First Steps Core Offline: • Fix the ESD Friends • Mods in AliReconstruction The Detectors: • Ask the detectors to evaluate number/type of events needed and size of data • Ask the detectors to write their AliAnalysisTasks C. Cheshkov & C. Zampolli
Any Comment? C. Cheshkov & C. Zampolli
Back-Ups C. Cheshkov & C. Zampolli
Offline Calibration – General Schema Chunk 0 Chunk 1 Chunk 2 Chunk 3 reconstruction ESD 0 ESD 1 ESD 2 ESD 3 i-th Run j-th Run Calibration/Alignment train Cal/Alig data 0 Cal/Alig data 1 Cal/Alig data 2 Cal/Aligdata 3 Merging procedure (*) Cal/Alig obj run in AliEn (*) analogous to the existing merging procedure for ESD tags C. Cheshkov & C. Zampolli