M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 1 CMS Software and Computing Outline n External (Common LHC, HEP) Software n CMS Software u Organization, Framework and Software Domains u Reconstruction and Detector Performance u Trigger u Fast Simulation n CMS Computing u CMS Computing TDR u LCG Service Challenge 3 u Software/Computing Integration n Major Milestones in CMS Software and Computing M. Stavrianakou On behalf of the CMS collaboration IEEE/NSS LHC Software: Crunch Time

External (common LHC/ HEP) software

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 3 External software currently used in CMS n Persistency framework: POOL n Core Libraries and Services: SEAL n Data analysis: ROOT n Generators: GENSER n Simulation: Geant4 n HEP class libraries: CLHEP, CLHEP/HepMC n Misc. libraries: BOOST n Parsers: Xerces n Visualization: Qt, Coin3D,… n Scripting: Python, Perl n Miscellaneous: u CppUnit, RuleChecker… u Oracle, MySQL…

CMS Software

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 5 Computing, Physics, Trigger (CPT) organization HLT/DQM E. Meschi Reconstruction T. Boccali Analysis Tools L. Lista Simulation M. Stavrianakou Calibr/alignm O. Buchmuller L. Lueking Framework L. Sexton Software L.Silvestris A.Yagil Computing L. Bauerdick D.Stickland Project Manager P.Sphicas Project Office V.Innocente L.Taylor Fast Simulation P. Janot ECAL/e-  C. Seez Y. Sirois TRACKER/b-  I.Tomalin F. Palla HCAL/JetMET J.Rohlf, C.Tully Muons N. Neumeister U.Gasparini Detector-PRS D.Acosta P.Sphicas Online Selection S. Dasu C. Leonidopoulos Higgs S. Nikitenko SUSY & BSM L. Pape M. Spiropulu Standard Model J. Mnich Heavy Ions B. Wyslouch Analysis-PRS A.DeRoeck P.Sphicas Generator Tools F. Moortgat S. Slabospitsky ORCA for PTDR S. Wynhoff Integration Program S. Belforte/I. Fisk Facilities and Infrastructure N. Sinanis Operations Program L. Barone Technical Program P.Elmer/S.Lacaprara SW Devel. Tools S. Argiro’

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 6 Software Project: two major lines n Physics TDR – Current thread u Goal: Maintain and support all current SW projects, in coordination with Physics Reconstruction and Selection (PRS) groups, until April/May 2006 n Migration thread u Goal: Software Components for the Cosmic Challenge l Framework (FW) l Event Filter and integration with DAQ and FW l Geometry DB/ Calibration and Alignment infrastructure (online/HLT/Offline) l Basic Reconstruction (Local) l Additional reconstruction needed for Muon Chambers l Detector Monitoring and Data Quality Monitoring l Basic Visualization n Cosmic Challenge will use new EDM and Framework u Essential test toward the integration of the full experiment

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 7 New Software: executive summary (I) n Framework in place n Event Filter (EvF) u Well tested in old FW; migration to new FW in progress n Calibration and Alignment u Calibration objects defined in collaboration with detector/reconstruction groups u First-round implementation of conditions service complete u Interval Of Validity (IOV) object prototyped u End-to-end (E2E) test well advanced

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 8 New Software: executive summary (II) n Simulation u Most components in place; integration for production testing scheduled mid-November 2005 n Geometry service in new FW u Done, being tested n Local Reconstruction u Well underway n Data Quality Monitoring (DQM) u In excellent shape; porting to new FW in progress n Visualization u Basic visualization for cosmic data in place

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 9 CMS data Excellent debugging tool real cosmic in HCAL at SX5

Reconstruction and detector performance

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 11 Electron reconstruction Effect of loosening seed cluster E T threshold and pixel matching on electrons in H→ZZ*→2e2μ Combining Ecal E and track P measurements for best 4-vector estimates (E scale corrections for Ecal, depending on e classification, full errors in weighted means) D. Futyan C. Charlot et al. H  ZZ*  4e M H = 150 GeV

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 12 Photon reconstruction Photon isolation studies Here looking for very large rejection factors needed for H  channel, and using combination of ECAL, HCAL and tracker T. Aaltonen, K. Lassila  signal rejection Previous result M. Pieri Photon error estimation Here shown as a function of shower width (sensitive to conversion, and conversion distance from ECAL)

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 13 Muons: reconstruction on cosmics data  MB3 sector 10 recHit 12 points track segments in one chamber  =510  m (expected from cosmics (not bunched!)

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 14 Z reconstruction reconstructed M(     ) 2 identified muons 1 identified muon, p T (2 nd track) > 10 GeV signal region 1 day 2x10 33 (events/ GeV) pp  WX bb  2  X HLT+offline cuts: P T 1,2 >20,10 GeV/c bb  2  X suitable for efficiency study from real data

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time GeV barrel HB cell Jet Measurement n A. Heister et al. AN 2005/05

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 16 HCAL response at low energy n test beam 2004 results J. Damgov et al.

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 17 Cosmic data of June 05 used in track-based alignment n Proof of principle: track-based alignment with CMS software tools works n Serves development platform for alignment algorithms n Future: large cosmic runs to validate TOB rod survey measurements n  Defines pre-data taking geometry! Mean track  2 Cosmic Rack / FinnCRack corresponds to a slice of TOB, 4 fully equipped rods were used Individual modules aligned First application of track based alignment using real CMS data convergence & good alignment in 5 iterations

Fast Simulation: FAMOS

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 19 FAMOS highlights FAMOS Number of Brem photons (>500MeV) for electrons with p T = 35 GeV/c  ORCA n Tracker material tuned layer by layer, as a function of , up to 1% accuracy wrt full geometry n Pileup simulation n Decays of long-lived particles (K0s, charged pions, …) n Muons: efficiency/resolution tables, muon isolation, global muons reconstruction n Electrons/Photons n HLT: JetMET, e/  and muons n Jets: shower parametrization, calibration, reconstruction… n Root-based analysis

CMS Computing

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 21 Computing TDR (C-TDR) n C-TDR goals u Summarize, refine the Computing Model (CM), expand on analysis and role of Tier-n centres u Explain the architecture of the CMS computing system u Detail the project organization and technical planning u Form the basis for further internal discussion n C-TDR - baseline of CMS computing services and workflows u Specify “baseline” targets and development strategy l Alongside LCG TDR, describing facilities and Grid baseline services u Not an engineering “blueprint” for the computing system n LHCC review of C-TDR ( ) successfully concluded

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 22 Computing Model Principles n Structured Data + Structured Data Grid u Primary Datasets, Data Tiers — optimizes sequential data access u Tiered Grid — facilitates well-defined roles/responsibilities for centers n Data Granularity — Primary Datasets u Data always needs to be considered in its trigger context -> trigger paths u O(2PB)/yr raw data split into O(50) (40TB) trigger-determined datasets n Data Tiers — RAW, RECO, AOD u Keep RAW and RECO close together (initially at least) —> FEVT u AOD, full copy at each Tier-1, partial (even full) copies at many Tier-2 n Computing Tiers u CMS-Tier0: prompt reconstruction, close connection to online u CMS-CAF: high-priority, short-latency data processing, analysis support u Tier-1s: FEVT data custody, selection, distribution, re-reco u Tier-2s: MC production, data analysis under physicist “control”

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 23 Technical Baseline Principles n Optimize for the common case: u Optimize for read access l Most data is write-once, read-many u Optimize for bulk processing, but without limiting single user n Decouple parts of the system: u Minimize job dependencies l Allow parts of the system to change while jobs are running u Site-local information stays site-local n Know what you did to get here: u ‘Provenance tracking’ is required to understand data origin n Keep it simple! n e.g.: initially use explicit data placement u Data does not move around in response to job submission u All data is placed at a site through explicit CMS policy

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 24 Computing Organization Principles n new computing organization based on projects and programs u realize the need for (sub-)projects that need to be well-scoped u require institutional responsibility to provide effort to projects n close coordination across management team u incl. weekly meetings w/ Grids: Integration Task Force, LCG PEB/MB n up-front: address the issue of “getting computing ready” u have computing model and technical baseline in hand u use integration milestones as Computing task drivers u system continually ‘up’, DC/SC, incrementally improving capabilities n establishing integration, (analysis support) operations, facilities, programs u a paradigm shift towards focusing on making things work end- to-end u together with WLCG, EGEE/LCG, OSG, CERN-IT, T1s, T2s et al.

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 25 LCG Service Challenge 3 (SC3) n Goal: demonstrate basic functionality expected at each computing tier simultaneously u Data can be transferred from CERN to Tier-1 sites u Data can be validated and existence of data can be published u Remote applications can be executed against local data to simulate skimming and selection jobs u Data can be transferred to Tier-2 centers, while validation and publication proceeds and remote analysis applications execute against data u At the same time, simulation jobs are executed at Tier-2 centers and results transferred to Tier-1 centers for archiving n Sites have validated CMS components for transfer during the throughput phase, publication tools are tested at most sites, grid interfaces are validated at all sites

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 26 Integration program n To ensure Computing delivers functional computing systems u preparing for the major service and data challenges u integrating CMS data management and workload management systems with EGEE and OSG middleware n Covers u Production-Related Activities u Analysis-Related Activities u Database Deployment Activities u Participation in Service Challenges

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 27 CMS-LCG Integration Task Force n Goal: Satisfy CMS needs making EGEE middleware work for CMS n “Make it work”: not stop at problems, fix them working with EGEE/LCG n TF is a CMS project, lead by CMS, part of Integration Program n TF is build around specific items that CMS needs n TF will not address all items in CMS agenda, just those that can be dealt with using EGEE middleware n TF will steer EGEE/LCG development and deployment to have them take over as much as possible of CMS needs

M. Stavrianakou CMS Software and Computing IEEE/NSS: LHC Software: Crunch Time 28 CPT: major milestones in 2005/06 n Computing TDR – done n Magnet test (cosmic challenge) u ready by Dec 2005; “slice” tests in progress u start taking data in January, for ~5 months n Physics TDR – Volume I u Draft 2 ready n Physics TDR – Volume II u Detailed outline ready n Data/computing/analysis challenge in 2006 u Currently the most important goal for 2006