LHCb computing in Russia

Ivan Korolko (ITEP Moscow)

Russia-CERN JWGC, October 2005

LHCb in numbersLHCb nominal Luminosity → 2x1032 cm-2 s-1

rate of p-p interactions → 2x107 per second

HLT output → 2000 Hz

RAW data per year → 2x1010 events (500 TB)

Events with b quarks → 105 per second (!)

acceptance for b-events → 5-10%

Br. for CP channels → ~10-5

Number of CP channels → ~50

~5 signal events in every second of LHCb operation – GREAT!

Have to select them from 1.5x107 background events

Signature of LHCb signals is not very bright (Pt , vertex)

Estimation of S/B ratio is a REAL CHALLENGE

LHCb computing documents

LHCb computing model LHCb 2004-119

LHCb computing TDR CERN/LHCC 2005-019

LHCb computing model (1)

1. RAW data → 2x1010 events/year

25 kB/event , 500 TB/year

1 copy stored at CERN and

1 copy in Tier1 centers

2. Reconstruction → 2.4 kSI2k.sec/event

2 times per year → 7 months and 2 months

CPU needs : 12MSI2k.years

storage: 500 TB (rDST data for stripping)

1 copy stored across CERN and Tier1 centers

4. Analysis → 0.3 kSI2k.s/event

140 physicists , 2 jobs/week , 3x106 events

CPU needs : 0.8 + 0.8x(n-1) MSI2k

storage: 200 TB

run at CERN, Tier1 and some Tier2 centres

LHCb computing model (2)

3. Stripping → 2.1x109 events/year

50-100 kB/event , 139 TB/year (DST)

4 times per year (1 month)

stored at CERN, Tier1 and some Tier2 centres

LHCb computing model (3)

5. MC production → 4x109 events/year

~50 kSI2k.sec/event (8 MSI2k.years)

400 kB/event

160 TB to store only triggered events (4x108)

produced at Tier2 centres

stored at CERN and Tier1 centres

LHCb trigger differs significantly from ATLAS and CMS

Russian participation in LHCb

IHEP (Protvino), INP (Novosibirsk), INR (Troitsk),

ITEP (Moscow), PNPI (St.Petersburg)

SPD and Preshower, ECAL, HCAL,

MUON system and RICH mirrors

Design, construction and maintenance of detectors

Development of reconstruction algorithms

Historical interests in B physics

Russian participation in LHCb

Physics interests of Russian scientists:

• Studies of CP violation with B and D hadrons

• Rare (forbidden) decays of B and D hadrons

• Radiative penguin decays pf B mesons (non SM physics)

• Spectroscopy of B hadrons

• High Pt jets (W and Z bosons, t quarks, Higgs, exotica)

Canonical users in Russian institutes:

IHEP -7, INP-5, INR-5, ITEP-7, PNPI-6

What to “compute” in Russia ?

There are 2 main tasks:

1. Provide facilities for LHCb data analysis in all participating Russian institutes.

2. Fulfill collaboration requirements

Computing in Russia (1)

1. Partially store reconstructed data (rDST)

most recent copy for the current year

1 copy for all previous years

storage: 225 + 225x(n-1) TB (b inclusive)

2. Store stripped data (DST)

most recent copy for the current year

1 copy for all previous years

storage: 140 + 140x(n-1) TB

Computing in Russia (2)

3. Run stripping jobs on stored rDST

B-inclusive sample – 9x109 events/year

CPU needs : 0.06 MSI2k.years

CPU power : 0.72 MSI2k.years (finish in one month)

data reduction factor ~100

TAG: 9 + 9x(n-1) TB

4 stripping passes per year → 0.3 MSI2k.years

rerun stripping during first years of LHCb operation

4. Run analysis jobs (20% of LHCb)

Computing in Russia (3)

CPU → 0.2 + 0.2x(n-1) MSI2k.years

storage: 50 + 50x(n-1) TB

MC analysis also requires some resources

B-exclusive dimuon D* B-inclusive

physicists 15 4 4 7

Computing in Russia (4)

5. MC production (~10% of LHCb)

CPU → 0.8 MSI2k.years

storage: 3 TB

6. Calibration, reconstruction algorithms

CPU → 0.2 MSI2k.years

storage: 30 TB

(hard to estimate now)

LHCb Tier2 cluster in Russia

CPU MSI2k disks TB tapes TB

2006 0.4 100

2007 1.0 300 100

2008 1.5 440 400

2009 2.0 820 500

2010 2.5 1200 600

partial Tier1 functionality (DST stripping, analysis)

History of LHCb DCs in Russia

2002 130K events, 1% contribution

only one centre (ITEP)

2003 1.3M events, 3% contribution

all 4 our centers (IHEP,ITEP,JINR,MSU)

2004 9.0M events, 5% contribution

started to use LCG

2005 PNPI and INR have joined

PlansMC production for the collaboration

routine task (running in 5 Russian centers)

participation in DC06

Distributed analysis

extensive tests of GANGA (CERN, Tier1)

deployment of GANGA in Russia

participation in SC

Significant increase of resources in 2006

learn how to operate much larger centers