1. G. Venanzoni-LNF KGM Frascati, 19 Dec 2006
Offline status report
G. Venanzoni-LNF
KGM
Frascati, 19 Dec 2006

2. Outline Status of the MC05 production Other offline activities in 2006
Offline plans for 2007
Manpower

3. KLOE MC upgrades for
Map of 2004/2005 machine and trigger conditions
New IR geometry
Better parameterizations of EmC response
Time, energy resolution; cluster efficiencies
Improved simulation of nuclear interactions/regeneration in DC wall and beam pipe
New event generators: KS  p+p-e+e-, h  p+p-e+e-, e+e-  wp0
Inserted background from events acquired with random trigger
Simulation of noise correlated to K tracks
Simulation of dE/dx measurement in DC
This “preparation” work ended in June 06

4. Status of the production for 2005 data
Pre-production test on a sample of 350 pb-1 successfully completed.
Real production started in August (on 104 Slots 218 B80 CPU) f  all events (LSF 1:1). From Oct 2 we were running on 136 slots (  280 B80 CPU)
Run Categories
Luminosity (pb-1)
Problems
(pb-1)
Good to be simulated
1123 -
Reconstr. (mcr)
1022 13
dst K± (mkc)
867 (85%)
167
dst K0 (mk0)
998 (98%) 36
dst rp (m3p)
1000 31
dst rad (mrn/mrc) 32
Dec 13 06
The inefficiency for mkc is mostly caused by the 2GB limit for single file!
We are missing (Dec 13) less than 100 pb-1. Will finish for Christmas ?!

5. Status of the production for 2005 data II
Dedicated work on data throughput to use 136 slots (  280 B80 CPU). More than 3 times of 2004/05 CPU power for MC.
# Slots
min
Aug 22
1day=1440 min
Aug 30
Sep 5
Sep 18
Oct 2
Oct 12
130
104
Time (min)

6. Status of the production for 2005 data III
136
24 Oct
27 Oct
8 Nov
10 Nov
136
23-27 Nov
15 Nov
13 Dec
In total  100 full days   10 pb-1/day on average
Still some improvement can be expected

7. Status of the production for 2005 data IIII
June/July:
many tests (and problems).
Mid of July->End of July:
datarec disks increased to 4+4 TB (from TB)
symmetric load server/machine
Simulation started but other errors encountered
August 22->Sept 12:
Real Start of the simulation with 104 slots.
Slots reduced time by time due to slow archiving rate (<<20 MB/s)
September 18 :
Restarted of the simulation, after fixing other problems (EMC cal banks, socket connection btw dbserver and fibm45,46,47)
Archiving rate increase to 18 MB/s on average
Oct 2:
duplicate archivers (one for each server). 132 slots.
Stable situation
Oct 16-24: shutdown, for server and disk upgrade. Restarted on 24 with 136 slots
Oct 25: Stop to fix an innocent bug in geanfi (omega pi routine)
Nov 8-10: Reduction of slots. Work done to improve the performances.
Nov 24-27: stop for disk cleaning and network operation
Nov27: Restarted. Smooth situation.

8. Offline shifters
20 people!

9. Monte Carlo production 2006
Program
Events
(109)
LSF
Time
(B80 days)
Size
(TB)
Simulation of f  all data
3.2 1
16920
96.4
All the Monte Carlo produced before 2006
Program
Events
(109)
Time
(B80 days)
Size
(TB)
Simulation of all 2001/02 data
2.3
11284
58.3

10. MC-data comparison for K±: E. De Lucia, P. De Simone
Data sample: 55 pb-1 of 2005 Data, K
Improvement in MC 2005 due to the simulation of correlated noise in first DC layers (SIMKBCK A_C module)

12. MC-data comparison for K0: M. Testa, K. Flowers

13. MC-data comparison for charged kaons: E. De Lucia, P. De Simone
ep hypothesis
data
data MC MC

14. MC-data comparison for charged kaons: E. De Lucia, P. De Simone
pp hypothesis
MC-data comparison for charged kaons: E. De Lucia, P. De Simone
data MC

15. 2005 data, 95 pb-1 KSp+ p-, KLp+ p-p0

19. Bug corrected in +-0 generator (thanks to F. Nguyen)

20. EmC time resolution in MC
C.Gatti
Three different contributions to the time resolution of EmC:
(T-R/c)= a ps E(GeV)½  b ps  c ps
Stochastic term
Constant term (uncorrelated)
[(T-R/c)]= a ps E*(GeV)½  2½ b ps
E*=E1E2/(E1+E2)
Common term (correlated)
(T-R/c) (ns)
E(MeV)
[(T-R/c)] (ns)
E*(MeV)
Time resolution is obtained for photons in a sample of +-0 events
both for data and MC and separately for Barrel and EndCaps.
Since the stochastic term cannot be easily fixed in the MC (10% difference in the barrel), we fit the data spectra by fixing the stochastic term to the MC value. In this way the difference is absorbed by the constant terms.

21. Data MC comparisons, after MC tuning: T-R/c Barrel

22. Data MC comparisons, after MC tuning: (T-R/c) Barrel
E=E1E2/(E1+E2)
data
data MC MC
Good agreement between data and MC both for barrel and EndCaps in all the energy spectrum considered ( MeV).

23. Distribution are simply normalized
Comparison data –MC e+e-→wp0 →p+p-p0p0 A. De Santis
Omega peak w p0 p r MC
Data
VMD diagram for
e+e-→wp0 →p+p-p0p0 MC
Data
KSKL peak
Data-MC(2005) comparison
Distribution are simply normalized
Good agreement

24. Other offline activities
Data quality for 2005 (km233) (S. Fiore)
Study of data quality for kpm stream (M. Martemianov)
HepDB to DB2 (F. Sborzacchi)

25. Data Quality (S.Fiore, G.V.)

26. Data quality tools: some updates
S. Fiore
CNDRUN.quality table being tested
New commands for data quality:
list_quality [run1][run2]
list_quality –long
list_quality –ok
list_runs -q
===============================================================================================================
RUNNR - insn - VERSION - K00PM - K00PM_REF - K00PM_S - DLUM - DLUM_REF - DLUM_S - SQRTS - SQRTS_REF - SQRTS_S
Default mode: displays run#, charged/neutral, vlab-trgmon, s, intlum
Displays all the quality infos + intlum, dead wires, hot wires
Same display as default mode, with charged/neutral and s within 3σ from reference, and intlum>20nb-1
Standard list_runs output, with the same requests of list_quality –ok

27. Study of data quality for kpm stream
M. Martemianov
Study of data quality for kpm stream
Analysed runs – (year 2005)
Used information from kpm – ybos files
For each run produced ntuples using 105 events
Double tagged of K→μυ events are selected

28. Kaon tracks / physics parameters / II
Charged kaons q – angle and momenta are quite stable
vs run number

29. Conclusion
Quality control of kpm – stream was performed on 2005 year data sample (around 80 runs)
Layer occupancy is quite stable and increased on ~10 %
Background parameters for kaon tracks (layers 1 – 12) are stable
Physical parameters for charged kaons is also stable
These quality controls will be done for the whole 2005 data sample and will be integrated with the information from list_quality command from S. Fiore and G. Venanzoni (memo n.331)

33. Offline/MC tasks for 2007
Reprocessing of 61pb-1 Off peak data (DBV-24)
Reprocessing and simulation of 2004 data: 645 pb-1
MC simulation Off peak data
Other business (MC generation for dedicated channels, reproc., etc…) ?
On the basis of the past experience,
we expect a rate  10 pb-1/day

34. What is need for the reprocessing and MC04?
DC maps Ready
Quality checks almost done
Physmon:
After run KeV were added
What to do for the 372pb-1 <32299? Re-do BMOM for all 372 pb-1, in order to have a consistent for all data?
How to treat the large time jitter period (run <30300) (see next slide) ?

35. (after data cleaning*) 507 pb-1 123 pb-1
Reason(s) not completely understood
123 pb-1
507 pb-1
30300 (21 Jul)

36. “Optimistic” time schedule
15 days for the preparation of the reprocessing 2006
10 days for the reprocessing 2006 (assuming 10pb-1/day (!))
1 Month for preparation of the reprocessing 2004
2.5 Months for the reprocessing 2004
2.5 Months for the MC 2004
TOT: 7 Months (without incl. contingencies…)

37. Offline resources: CPU
Nodes
Type
CPUs
B80
“MC”
“CW”
fibm14-15,17-34
P3 375 MHz 80
fibm35-44
P GHz 40 88 96
fibm45-47 (Jan ’06)
P5 1.5 GHz SMT
“96”
126
184
All
294
360
KLOE CPU accounting is performed in B80 units
1 “B80 CPU” = 1 P3 CPU installed in B80 server
Uncertainties transferred to B80/P4+ and B80/P5 ratios
“MC” estimates based on MC tests conducted May 2006
“CW” estimates based on past experience

38. Offline resources: Disk space
DSTs cached on nfs-mounted disks for fast analysis access
DST volume*
Current
Final
35 TB data
24 TB MC
59 TB total
42 TB data
44 TB MC
86 TB total
*Updated From M.M. (Sci.Com.)
~60 TB available for disk cache
Current DST cache capacity: 14 TB
New purchases: 21 TB FC disk available
Installed other 30TB Sata2

39. Tape Storage (Dec 06) Tot 640 TB
Data 2001/02 (TB)
RAW
101
Recon 55
DST data
7.5
Total
163
Luminosity Tag=OK
440
2004
677
2005
1233
Scan+OP
282
Total
2632
175 TB MC
Data 2004 (TB)
RAW 54
Recon 40
DST 6
Total
100
Data 2005/06 (TB)
RAW 95
Recon 86
DST 21
Total
202
Tot 640 TB
Quanto assumiamo per la luminosita’ 2001:
Capacity of tape library with current cartridges: 930 TB ( )

40. Manpower Seems the weakest point:
2 FTE + 1 fellowship (end March 07) for the Computing.
1.5 FTE (3 persons) + 20(!) Offline shifters for the offline.
Optimistics scenario for 2007:
Manpower improves….Seems not the case (even if we always hope)
Realistic scenario:
manpower will continue as it is now
Unsunstainable scenario:
The Fellowship will not be renewed….

41. Outlook and summary
MC simulation for 2005 data almost finished. A huge effort (for) from many people!
Offline activity will continue to be LARGE in 2007
Manpower can became a critical factor