offline status report
DESCRIPTION
Offline Status Report. M. Moulson, 27 Feb 2001 Summary presentation for KLOE General Meeting. Outline: Processing of 2000 data Upcoming developments for datarec executable Status of MC production Ntuples/DST’s: status and proposals. Processing of fall 2000 data. Reference data set: - PowerPoint PPT PresentationTRANSCRIPT
Offline Status Report
M. Moulson, 27 Feb 2001
Summary presentation for KLOE General Meeting
Outline:
• Processing of 2000 data
• Upcoming developments for datarec executable
• Status of MC production
• Ntuples/DST’s: status and proposals
Processing of fall 2000 data
Reference data set: Runs 15197—17330 (23 Sep—11 Dec)5.61G triggers, 15543 files, 13.6 TB raw data
Offline L dt = 23.1 pb-1
• DBV-6: 5.5 pb-1
• DBV-7: 18.0 pb-1
Total events collected:
• 130M Bhabhas
• 67M decays est. = 2.9 b
• 10.9M KL tags
• 7.2M KL crash
• 19.5M KK w/ vertex
• 6.2M tags
Processing statistics for fall 2000 data
Raw data: 5605M Events, 13.6 TB
Physics: 2.0% of total events25% of raw volume160 GB/pb-1
Bhabhas: 2.3% of total events20% of raw volume114 GB/pb-1
Stream Events (M) Volume (GB) Summarykpm
ksl
rpi
rad
clb
19
64
6
23
1
687
2270
166
520
19
112M Events
3.6 TB
bha 127 2623flt
afl
cos
1248
31
96
3791
92
273
1379M Events
4.2 TB
Avg. fraction ofevents tracked:
26%
Avg. bandwidth on 40 CPU’s: 1060 Hz
Avg. time/trigger
38 ms
Avg. time/tracked evt.
151 ms
datarec performance in fall 2000Daily averages
Average uptime efficiency: 63%Staging and procedural losses count
Evts. acquired Evts. processed
Once in stable situation, able to follow with ~75% of data processed
No
DC
trig
End
of
data
Fol
low
-up
Aft
er X
mas
Mis
sed
runs
Rep
roce
ssin
g
Acquistion rateReconstruction rate
Farm bandwidth
23 Sep 11 Dec
Data processing: Issues to discussIntended to give feeling
Not an exhaustive list
More detail at review
Reprocess DBV-6 (5.5 pb-1) for neutral radiatives?
FILFO Bhabha filter deleteriousRejects 40% of events
Only runs on “charged” eventsPre DBV-7: Charged event: Nhits > 30Pileup contribution to Nhits in 2000 data
DBV-7: Bhabha filter vetoed by neutral radiative tag instead of Nhits cut
Luminosity to 1% only possible with DBV-7 reconstruction
Streaming-related issues:
flt stream (raw)About 30% of raw volume (3.8 TB in 2000)Apparently never used
selcos cosmics10% written as raw, also processedIncrease statistics in raw, don’t processIncrease number processed, don’t write raw?
Reprocess runs processed before final EmC calibrations available?
Currently we do not wait for calibrations before reconstructing
Complete EmC reconstruction is performed before Ntuple/DST production
Ntuples/DST’s, YBOS out of phase
Effects on streaming are permanent
Next version of datarec executable
Already in DBV-8:
CLUFIXENE MC events, scale factor
EMCALIB slewing corrections
VTXFIN corrected path length calculation
TRCLMOD modified projection from track to cluster
Ready for inclusion in datarec:
FILFO fast tracking cuts, Nhits definition
VLAB new luminosity module
A_C zlib event compression
MC: f events
Before scale corr.
After scale corr.
CLUFIXENEHow it works:
• loop over “missing” cells• attenuation length correction from zcl
• sum EA and/or EB into cluster
fscale because ABSCAL done w/o CLUFIXENE
Problems:• CWRK not updated for missing cells• Ecorr Ecl – Ecl
bef not saved in bank stream
• Makes fscale necessaryABSCAL uses CWRK
New developments:• CLUFIXENE now works for MC events
• fscale off by 2%—adjusted
scalemiss cellbefclcl fEEE
In progress:• Bookkeeping mechanism to replace CWRK
CSPS bank includes missing cells • ABSCAL will run on CSPS
Remove scale factor
EMCALIB: Slewing corrections
TcorrA,B = TA,B – T0
A,B – TslewA,B
TslewA,B = Const/(EA,B)0.1
• Empirical prescription, same for all cells
• Tested on “old” data with T0’s adjusted by hand
• Correction now applied in T0ALIG and TIMCAL
• Disabled in code for runs < 17606
Diagnostics:•T1T5 distributions OK
•v12 distributions OK
•FILFO should not be affected
T5
T1
v12
Slewing correction now applied cell-by-cellee evts. w/w/o slewing correction
ps 140)GeV(
ps 60
ETT
CLUREC: New clustering cuts
Most split clusters are actually unmerged clusters
Basic problems:• Current cuts do not depend on Ecl
• Complications from shower fragments
Problem of unmerged clusters has been comprehensively studied
ee SM + LF
CD
f SG
KSKLKSKLcrash GC
KSKLKSKL MC
r, r, z 6 cm/E(GeV)
t 3.5 ns
Attempt to tune CLUREC cuts to minimize cluster splitting
t 2.5 ns
z 30 cm
r 20 cm All four conditions must be met
r 40 cm
Clusters are merged if:
Proposed new merging criteria:
Status:Performance gains to be evaluatedImplementation easy
0.8
1.0
1.2
100 200 300 MeV
Cluster efficiency studies
e
e
From ee events:• Good vertex, 2 tracks,
high p
• p1, p2 > 490 MeV• Exclusion region around
e± clusters• pmiss(DC) gives direction
of expected • Systematic from cone
definition• Studied in data, MC
From events:• Good vertex, 2 tracks• pmiss(DC) = p
• 1 direction gives expected E1
• Compatible 1 identified direction of 2
• Systematic from cone definition• Advantages:
higher pmiss(DC)smaller systematic with , E
MC Data
MCData
Tail normalized
recMC
recdatareccell EN
ENER
Ecell (MeV)
from ee evts. with thresh. simulation
MC Data
Cluster efficiency: Data and MCSimulation of cell thresholds
MC: • Only when no p.e. after attenuation
simulated at one side• Relatively hard cutoff at ~2 MeV
Real life:• Roll-off, characteristic value ~7 MeV
Compare MC to data using ee eventsErec for all cells associated to clusters
on barrel, Ecl > 200 MeVsimilar shape on endcaps
A_C module EMCFAKETHR:• Deletes CELE hits from MC events according to
hit-or-miss sampling of Rcell
• Brings (E) for data and MC into agreement for high
• Some systematics to be studied:+1 MeV offset in Erec
MC to obtain good results
• Ultimately hope to avoid having to map Emethod
to EKINE when applying corrections
Status: almost ready for release
EmC data structuresCTRG: Sector cluster correspondence bank
Combines information from PZZA, TELE banks• Number/kind of sectors not connected to clusters• For all clusters: indices/kind of sectors incorporated• Can also obtain indices of clusters in each sector
Status: done, also in PROD2NTU
CSPS: Calorimeter space-point bankCombines information from
• CELE: EA, EB, TA, TB
• CWRK: (E, X, Y, Z, T)cell
into one compact structure, capable of storing multiple hits per cellStatus: development, almost ready
CLUS: Compact cluster bankAnalogous to CSPSStatus: ideation
FILFO: Cuts and DC noise
FILFO cuts on number of DC hits (Nhits)
Cosmic rejection:• By crossing speed, shower depth cuts for Nhits < 175• By ring cuts for 0 < Nhits < 60• Cosmics discriminated from KK events by Nhits < 100
under certain conditions
MB rejection: • MB events have Nhits < 200• Min EmC energy vs. Nclust for neutral events (Nhits < 50)
Noise hits in DC create interferenceHot channels masked in RAW2DTCE
Problem from diffuse component of noise• Does not depend on detector• Can be estimated from Nhits(t < 0)
Nhits(adjusted) = Nhits(raw) – Nfake
measured estimated
0 t
Nh i
ts
Status: RAW2DTCE already modifiedFILFO modification trivialGains under study
FILFO: Fast trackingEmC reconstruction
t0 step zero
ATFMOD/VTXFINpar = 1
FILFO
FILFOTRK
ATFMOD/VTXFINpar = 4
ATFMOD/VTXFINpar = 3
Event Classification
.bhaxxx
RAW
Mach. Bkg.Cosmics
Single trackevents
UFO’s
Physicsstreams
Bhabhastream
Modifications to ATFMODAdded parameter suppresses:
• Hit addition/rejection• Kink finding• Multiple scattering• Number of iterations during refitting
Similar procedure used for Bhabhas
Reduces reconstruction time 4—5
New module: FILFOTRKRejection based on preliminary tracks
• Currently implemented:Single track events
• Eventually:Inclusive cuts based onevent selection algorithms
Status: Code structure completeGains fully realized with new rejection criteriaPossible to jump-start tracking?
VTXFIN: Path length calculationErrors fixed in calculation of track length from FH/LH to vertex
Vertex position not affected
2
1Relative to analytic (helix) calculation:Should have some tails
Stepping error (1)Avg. effect: 2.5 cm (modularity: 5 cm)
Numerical extrapolation of trackUses B, takes into account dE/dx5 cm step size (except last step)
For L, vertex was put at beginning/end of step because calculation was incorrectly started from FH LH
Endpoint error (2)Avg. effect: 20 cm
Vertices b/t BP and DCTracks which intersect BP
Both fixed in DBV-7
Function vtxswm_p2nAnalytic calculation of L
Works with PROD2NTU
BP
DC20 cm
OK
TRCLMOD: Recent developments1. Track length from LH to EmC systematically overestimated by 2cm
Pre DBV-8: Can simply subtract 2 cm from this lengthOften canceled with –2.5 cm error in VTXFIN track length (hard to detect!)
2. Changes in calculation of association
EmC
OLD
NEW Old cuts:, z (barrel)x, y (ecaps)
New cut:Distance
Cuts are large—does not have a large effect except for case of large angle of incidence
3. Talk-to option to force TCA for tracks that end at vertex
TCA usually inhibited—idea is to follow “tree”
May recover cases involving split tracks
Not default behaviorContemplate use in KS semileptonic analysis in order to get better control over TCA efficiency
Luminosity measurement in 2000Traditional offline L estimate from LAB’s:
• 2 clusters with 45° < < 135°• 300 < E < 600 MeV• Acollinearity in < 10°• In time within 4 ns• Acollinearity in < 80°
Problems:• separation: 50 hits in DC
Up to 8% background from ’sesp. at beginning of 2000 running
• Emax cut unstable w.r.t. energy calibration(E) = 42 MeV at E = 510 MeV3% effect
On-the-fly fixes in 2000:• Exclusive 3D acollinearity cut to
eliminate ’s • Emax 600 800 MeV
3dim - Accolinearity3dim - Accolinearity21
21cosClCl
ClCl
xx
xx
)( eeee
)(ee
New Version of BABAYAGA !
cos > -0.975
1%
VLAB luminosity measurementStarting sample: FILFO LAB’s
with 2000 cut revisions
Event selection:2 tracks from IP with opposite chargerxy < 7.5 cm, rz < 15 cmDistance between 1st hits > 50 cm
VLAB cuts:55° < (cl) < 125°Acollinearity in < 9°400 MeV < E(tr) < 600 MeV
MC effective cross section = 421 nb0.3% background from events manually subtracted
New L estimate from events (GGL)FILFO LAB cutsCut on Nhits vs. 3D acollinearityMC effective cross section = 118 nb
GEANFIBABAYAGA
Data
E + -
+ -
VLAB status
VLAB routine is well tested and performs as expected
• estimated systematic effects from cuts on E, : L/L < 2%• best-tested and most accurate luminosity measurement yet
VLAB has been implemented in an Analysis Control module• all relevant parameters accessible via Talk_To
• will be inserted in next datarec version
• long term method test by comparison to Gold Bhabhas, LAB’s, and GGL
Number of VLAB candidates to be written to DB on a file-by-file basis• statistical accuracy for a single run: 30 nb-1 L/L= 0.9 %
Plan to process DBV-7 bha stream to produce luminosity map (17 pb-1)
• Pre DBV-7 executable had Emax(cl) = 600 MeV cut LAB sample biased
New version of GEANFIGEANFI version 1.06/07 released on 6 February:• EVA generator inserted
generator with different models for ISR, FSR, interference with f0• decay inserted• decay (C-violating) inserted• Improved simulation of K 3 decays generator
added parameter to specify fraction of decays that don’t proceed via • Modifications to GEANT tracking to allow generation of specific KL topologies
KL decays in beam pipebefore entering DCbefore entering calorimeterin calorimeter
• Change name of output banks: DTCE MDTC, CELE MCELMakes it easier to re-reconstruct MC files that have already been reconstructed
• New default value for magnetic field: I = 2300 A
Modified MC banks and analysis chain
RAW2CELE/RAW2DTCE
All other reconstruction
TSKT/SMEAR_T0
T0_FIND/CLUFIXTIM/T0_MENO
EMCALIB/CLUREC/CLUFIXENE
mcr
mco MCELMDTC
CELE DTCE
Smeared banks
Corrected banks
MC
EL
, MD
TC
RAW2CELE/RAW2DTCE
All other reconstruction
T0_FIND/CLUFIXTIM/T0_MENO
EMCALIB/CLUREC/CLUFIXENE
xxx
raw CALRDCHR
CELE DTCE
Corrected banks
CA
LR
, DC
HR
MC Data
• New names for MC output banks:CELE MCEL, DTCE MDTCPristine banks for re-reconstruction
• New A_C path for MC reconstruction:
use (dmapload) emcdbini raw2cele dcdbini raw2dtce (tskt smear_t0) emcalib clurec clufixene t0_find clufixtim t0_meno filfo dconvr…
• For MC versions 167 and later• Re-reconstruction w/w/o redoing
trigger simulation/t0-smearing OK
• For MC versions pre-167:
• Original trigger simulation/smearing automatically used
• Re-reconstruction w/ new trigger simulation/t0 smearing not possible
• New routine to obtain MC versionGET_MC_VERSION(VERSION)
Status of MC production
Event type Request 26 Feb
all 2.4M 2.4M
KSKL
KS ; KL various
KS ; KL all (1.1M)
2.9M 1.3M
KK
K ,e0.24M 0
radiative channels,,fa
ee
2.0M 1.7M
calibration channelseeee
ee
cosmic rays
0.8M 0
Total ~8.3M 5.4M
Production details:
New GEANFI version (1.06/07)
New datarec version (DBV-8)
8 offline nodes (fsun03-10)
Environment similar to datarec
Output:
Reconstructed YBOS files (.mcr) archived to tape
Program:
Continue with MC production until farm bandwidth needed to reconstruct new data
9 days at current rate
MC production for 2000 data started 9 Feb
Ntuple production for ksl streamStarts from ksl stream
• EmC reconstruction• t0 find (step 1)• PROD2NTU ksl Ntuples• ksl Ntuples split into substreams• YBOS output for KL events• manual tape backup
Ntuple Tags
kslall events in ksl stream
ks2piKS
KS (gold)
klcr KL crash
ksemilKS semileptonic decays
kl3p0 KL
klppp KL
intertag interferometry tags
Status:
About 4 pb-1 processed in Jan
23 GB of Ntuples!• Eliminated KS tag• Dropped DPRS block• TELE, PIZZA blocks condensed
into C2TRG
Restarted 19 Feb, immediately stopped • Initially for machine problems• Fine-tuning of scripts, output file
sizes• 30-50% reduction in data volume
Restarted this weekend, 1 pb-1 processed
DST production for radiative streamStarts from rad stream
• EmC reconstruction
• t0 find (step 1 radiative)
• Retracking for events that overlap with bha stream
• YBOS DST’s:
• 4 substreams for analyses in progress
• Subset of events in rad stream
E.g., ee, events
• Reduced bank set
• Manual tape backup
• Diagnostic Ntuples for some substreams
Output stream Topologies
neu_max4gneutral events, at most 4 prompt ’s
neu_min5gneutral events, 5 or more prompt ’s
cha_min3g2 tracks3 or more prompt ’s
cha_ppg tagged events
Diagnostic Ntuples:Allow estimates of events/nb-1
Status:
•All data reconstructed with DBV-7 completed
•Total size of all YBOS output: 40GB circa
Outlook for mini-DST production
Proposed format for mini-DST’s
•zlib-compressed YBOS
•Bank subset essential for physics analysis
•Summary banks?
•Substreams?
Phase 1:
Mini-DST production in dedicated pass
Phase 2:
Mini-DST production as part of data processing
Phase 1 Status:
Compression DoneScripts Some work neededRetained banks Needs definitionSummary banks Needs definition, codingSubstreaming Needs definition
Waiting on physics inputTechnical aspects relatively simple
Phase 2 Outlook: More distantTechnically more complicated
E.g.: what to do when files from a given run processed weeks apart?
Solutions exist, but limited resources to devote(man and machine)