Date post: | 06-Jan-2018 |
Category: |
Documents |
Upload: | sophia-williamson |
View: | 228 times |
Download: | 2 times |
27 June 2006Shuei YAMADA, MEG review @ PSI 1
MEG SoftwareStatus and Prospects
MEG Software group
27 June 2006Shuei YAMADA, MEG review @ PSI 2
Contents
Software documentationMC updates WF simulation and analysis tool updates ComputingMEG software releaseMEG data challenge
27 June 2006Shuei YAMADA, MEG review @ PSI 3
Software documentation
27 June 2006Shuei YAMADA, MEG review @ PSI 4
InstallationMC introduction
User’s guide
FAQ
Developer’sguide Source code
browsing
MEGWiki
27 June 2006Shuei YAMADA, MEG review @ PSI 5
Document generation
HTML docs are generated from C++ source codeGuide lines to write comment : http://root.cern.ch/root/Conventions.htmlGenerated docs from the HEAD version available on :
http://meg.web.psi.ch/restricted/software/meganalyzer_htmldoc/ClassIndex.html andhttp://meg.web.psi.ch/restricted/software/meganalyzer_htmldoc/ClassIndex.html
Sourc
eHTM
L
27 June 2006Shuei YAMADA, MEG review @ PSI 6
MC updates
27 June 2006Shuei YAMADA, MEG review @ PSI 7
Event generationMore calibrationsRadiative decayCosmic ray (preliminary)
More precise DCH hitsTICP/TICZ output to be digitizedUpdates and fixes in geometries
LXe outer wall : 846 PMTs in totalUpstream/downstream beam elementsDCH dimensionTICZ/TICP
TC beam test MC
MC updates
27 June 2006Shuei YAMADA, MEG review @ PSI 8
Items ready for productionEvent generator
Signale+ + from radiative decaye+ from Michel decaye+ from radiative decay from radiative decayCalibrations
AlphaNie+e- pair from W converter
Cosmic ray (preliminary)ZEBRA output to be digitized in megbartender
All the sub-detectors are ready: LXe, DC, TC(Phi), TC(Z)Preliminary trigger simulation
27 June 2006Shuei YAMADA, MEG review @ PSI 9
Good level of sophistication, but items to be done still exist…Event generator
AIF Calibrations
LEDLi(p,)0
Cosmic rayPreliminary – zenith angle distribution is readyTo be done : azimuth angle distribution, energy spectrum
Neutron backgrounds for LXeBackground from beam
Crosscheck with beam testTiming counterDrift chamber
27 June 2006Shuei YAMADA, MEG review @ PSI 10
Waveform simulation and analysis tool updates
27 June 2006Shuei YAMADA, MEG review @ PSI 11
Event display
More displays : tracks and energy deposits
27 June 2006Shuei YAMADA, MEG review @ PSI 12
Xe Waveform display
Expected To be used also for online
27 June 2006Shuei YAMADA, MEG review @ PSI 13
DCH 2D display
27 June 2006Shuei YAMADA, MEG review @ PSI 14
( )
MEGWaveform class
Reorganization of waveform analysis/simulation classEach sub-detector/analysis/simulation was using its own waveform class
MEGWaveformMEGDRSWaveform
MEGTriggerWaveform
MEGXECDRSWaveform
MEGXECTriggerWaveform
Basic operationsadding/subtracingDrawPrint…
Detector independent operationsADC/TDCFrequency filteringReflectionMoving averageSplitter responseDRS simulation…
Detector specific operationsIf neededPMT/APD/Wire/Pad sim.Fitting…
27 June 2006Shuei YAMADA, MEG review @ PSI 15
Standard system of unitsDiscussed and decided the standard units for MC and offline analysisDefines internal expression in programs
constant valuesglobal variable and Fortran common blocksreturn value and arguments of functions/subroutines/methodsData file (ZEBRA, MIDAS, ROOT, …)
Any unit for human interfacehistograms, plotsConfiguration in GUI panelWarning messages
Units based on GEANT3
Base units * Energy * Time * Length * Angle * Voltage * Temperature
GeVsecondcentimeterdegreeVKelvin
27 June 2006Shuei YAMADA, MEG review @ PSI 16
Xe analysis updates & schedules
UpdatesLinear fit & MINUIT fit reconstruction were implemented (by Giovanni)Pattern recognition (both time and space) for pile-up events (by Yasuko)
To be doneQsum-based reconstructionWaveform spectrum analysis in progressTiming reconstructionCalibration & monitoring w/ alpha, LED, pi0, proton, RD, and neutrion
ScheduleFinish items listed above before starting MC mass productionValidate analysis tools in August-September
27 June 2006Shuei YAMADA, MEG review @ PSI 17
DC offline updates and To-dos
AnalyzerKalman filter reconstructionCOBRA 3D B-field based on measurement
BartenderImplemented most physical effects in wire signal simulation
Drift electron arrival times, avalanche and current pulse shape, charge division, preamp, cable propagation and reflection, DRS responseTo be checked against 55Fe Source
To be donePad simulation was just started
Precise charge integration of induced chargeSignal electronic transformationCompare analytic formula for reflected signal w/ measured data…
27 June 2006Shuei YAMADA, MEG review @ PSI 18
DC Wire output
right
left
27 June 2006Shuei YAMADA, MEG review @ PSI 19
Status of TCAnalysis
Very preliminary stage - No algorithms implementedAnalysis of test beams (@Frascati & @PSI) are underway MC should be validated
BartenderTC(Phi) waveform - some physical effects implemented
Need tuning and corrections to mach with electronicsCode copied from LXe – differences in electronics chain
TC(Z) waveform simulation – not started yet
New TC organizationCoordination : P.W. CattaneoMEG MC simulation (GEM): S. DussoniMEG Digitization (Bartender): A. BarchiesiMEG Analysis (Analyzer): P.W. Cattaneo, D. ZanelloTest Beam MC simulation: S. DussoniTest Beam Analysis : A. Barchiesi, D. Zanello
27 June 2006Shuei YAMADA, MEG review @ PSI 20
Status of trigger simulationμeγ mode partly implemented
Missing itemsTICP waveformTICZ waveform
Coming soonXe trigger waveform and discriminationXe and TC timing matching ( Earliest TICP hit is used as TC timing )Xe and TC direction matching ( Earliest TICP hit is used as TC position )
Should be replaced with PMT/APD output
27 June 2006Shuei YAMADA, MEG review @ PSI 21
Computing
27 June 2006Shuei YAMADA, MEG review @ PSI 22
Merlin3 cluster @ PSIMerlin3 : 54 Cores + 14TBytes Disk + batch (SGE : Sun Grid Engine)No modification needed for MEG software
Write a script for SGE and submit a job
Potentiality of use of GRIDNo modification needed for MEG softwareSuccessfully tested for MC production (only 10 events)GRID access is limited to Virtual Organizations (VOs) : cms, atlas, gilda, … Need more tests
#!/bin/bash### Change to the current working directory:#$ -cwd### Job name: #$ -N simple_arrayMY_HOST=`hostname`MY_DATE=`date`echo "Running on $MY_HOST at $MY_DATE"echo "Running environment:"envecho "========================================"# Put your single-CPU script here./Linux/gem gem.cards
27 June 2006Shuei YAMADA, MEG review @ PSI 23
Offline cluster for MEG
Ordered on June 14th : 20 cores + 30 TB disk※ MEG requirements : 64 cores + 100 TB disk in totalEasily extensibleRedundancy through GFS/GPFS file systemsGBit link to online cluster requested
Sun Fire x4100 quad core 4 GB
Sun Fire x4100 quad core 4 GB
Sun Fire x4100 quad core 4 GB
Sun Fire x4100 quad core 4 GB
Sun Fire x4100 quad core 4 GB
Fibe
r Cha
nnel
Swi
tch
15 x 500 GB SATA
GBit
Ethe
rnet
27 June 2006Shuei YAMADA, MEG review @ PSI 24
DatabaseTwo databases : MySQL and sqlite3
MySQL needs networksqlite3 for stand-alone environmentEasy conversion from MySQL to sqlite3
Test database is available inside PSIRun tableLXe
Geometrycable configurationPMT test resultLinear-fit parameter
Requested tables for TC, DC and DRSDiscussions continuing on ELOG/Offline
DB mirroringMC run tableTrigger configuration
27 June 2006Shuei YAMADA, MEG review @ PSI 25
MEG software release
27 June 2006Shuei YAMADA, MEG review @ PSI 26
On June 19th: meg-1.0.0 released
Released even though couple items left behind- We cannot wait for everything being readyPerform MEG Data Challenge
Feedbacks are welcomeBug reportsRequestsQuestions…
Next release (meg-1.1.0 ?) To be used for MC mass productionRelease by the end of July
Start MC mass production from August
27 June 2006Shuei YAMADA, MEG review @ PSI 27
Release strategy
Release independent of MEG
ROME Online
MEGMCMEGBartenderMEGAnalyzer
MEGMEG
RepositoryROME Repository
All the MEG software are tagged togetherRelease early, release often
When detector setup changedWhen data structure changedWhen interface changed
ROME, ROOT, CERNLIB, MySQL, etc…Some amount of bug fixes/new feature accumulated
Accumulate for one month? Announcement posted ~ 1week in advance
Responsible persons have to make sure the subproject is in a good shape at that date
27 June 2006Shuei YAMADA, MEG review @ PSI 28
Retaining compatibility :Data structure
Compatibilities btw. MC, Bartender and AnalyzerMost probable point to screw up the analysisChanging data structure implies changes to all code using that data
Make sure that newer Bartender/Analyzer can read old MC dataunrealistic to perform MC production for every different data structure Append new data into ZEBRA is less problematicRemoving from/changing order of/inserting data ZEBRA are problematic
MCA⇒B
bartenderA⇒B
analyzerA⇒B
27 June 2006Shuei YAMADA, MEG review @ PSI 29
MEG data challenge
27 June 2006Shuei YAMADA, MEG review @ PSI 30
MEG data challengeTest throughout MC production, Bartender, and Analyzer
Test meg-1.0.0 : find out unknown missing items, problems, …Realistic estimate of CPU/DiskASAP
Perform before MC mass productionStart with event samples available
Signal, e+ from Michel, RD, Calibrations, CR, …104-105 events/event type
~ 1.5GBytes/104event for signal event~ 1day/104event for signal event
Available resources at moment : merlin3 cluster @ PSICPU : 4core Opterons x14, 2.4GHzDisk : 14TBytes
27 June 2006Shuei YAMADA, MEG review @ PSI 31
End of Slides
27 June 2006Shuei YAMADA, MEG review @ PSI 32
Dataμeγ trigger 2Hz * 20days = 4M events 〜 3TBCalibration
Alpha, LED : 200MB/run * 4times/day 〜 50GBPi0, RD, neutron, protonDAQ w/o LXe in late summer
Need discussion for the amount of calibration data(10Hz * 20days = 15TB, too big !!)
MCRD 4M event 〜 4.5TB Michel 1M event 〜 1TB Michel 1M event 〜 1TB μeγ 1M events 〜 1TBCalibration 2TB
Storage needed for this year
( w/o waveform )
3 TB+
calibration
( low intensity )
10 TB 20 TB ?
27 June 2006Shuei YAMADA, MEG review @ PSI 33
CPU estimateMC production + digitizing > 20 CPUs
6-8 seconds/event * 9M event = 625-833 daysLow intensity beam : single RD is dominant
Waveform fitting 2 CPUsfitting 0.5msec*600 wf/event = 0.3sec/event0.3 sec/event * 1.2 M event = 1.4 daysReduction of 1/10 by pre-selection assumed
Analysis 6 CPUsParameter training (Linear fit)CalibrationPre-selectionWaveform fast analysis (T,Q conversion) 0.5 event/secReconstruction 0.5 sec/event ?
To be consideredTime for submitting job – negligible?Time for tape accessMC and analysis do not run simultaneously