event processing applications summarycomputing in high energy physics 2006 – mumbai (india) event...

Event Processing Applications summaryComputing in High Energy Physics 2006 – Mumbai (India)

Event Processing Event Processing ApplicationsApplications

summarysummary

Gabriele Cosmo (CERN)Gabriele Cosmo (CERN)Co-convener: Avi Yagil (FNAL)Co-convener: Avi Yagil (FNAL)

Chairs: Witold Pokorski, Alberto Ribon, Liz Sexton-Kennedy Pedro Arce, John Apostolakis, Chris Jones

17 February 200617 February 2006 Event Processing Applications summaryEvent Processing Applications summary 22

OutlineOutline Contributions Application Areas

• Simulation In the experiments Simulation tools Simulation validation

• Frameworks & data models• Reconstruction

Tracking algorithms Alignment

• Visualization, commissioning Impressions …


ContributionsContributions

40 parallel talks40 parallel talks• Not all talks treated here …Not all talks treated here …• All slides are available on the webAll slides are available on the web

25 posters25 posters• Not treated in this talkNot treated in this talk

44

PresentationsPresentations


SimulationSimulation

Geant4 simulation• In production: CMS, ATLAS, LHCb• ILC, BESIII, CBM• Heavy ion therapy

Fluka• The future main simulation engine in ALICE

(Geant4 as alternative)• Source code available on request as of last July

2005 release Need for fast-simulations

• CMS, ATLAS


#3 – A.Rimoldi, The ATLAS Detector Simulation – an LHC challenge

77

CMS simulationCMS simulation Switched from GEANT3 to GEANT4 Switched from GEANT3 to GEANT4

simulation end 2003 simulation end 2003 So far the CMS OSCAR simulation So far the CMS OSCAR simulation

package delivered ~100 M physics package delivered ~100 M physics events for CMS 2004 Data events for CMS 2004 Data Challenge and on-going Physics Challenge and on-going Physics TDR studies TDR studies

OSCAR being adiabatically replaced OSCAR being adiabatically replaced by a new suite (SimG4), based on the by a new suite (SimG4), based on the new Event Data Model Software new Event Data Model Software FrameworkFramework

CPU: SimG4 CPU: SimG4 1.5 x SimG3 1.5 x SimG3 with lower production cuts!with lower production cuts!

Memory: ~110 MB/evt for pp in Memory: ~110 MB/evt for pp in SimG4SimG4 vs. ~100 MB in SimG3vs. ~100 MB in SimG3

Robustness: from ~1/10Robustness: from ~1/1044 crashes in crashes in pp events in DC04 to < 1/10pp events in DC04 to < 1/1066 crashes in latest productionscrashes in latest productions

#243 – D.Elvira, M.Stavrianakou, CMS Detector and Physics Simulation

88

ALICE ALICE simulationsimulation

FLUKA VMC FLUKA VMC implementation completedimplementation completed

Testing well advancedTesting well advanced• TTGeoGeo/FLUKA validation /FLUKA validation

completedcompleted• Good agreement with G3 Good agreement with G3

and Testbeamand Testbeam FLUKA VMC will be used in FLUKA VMC will be used in

the next ALICE Physics the next ALICE Physics data challengedata challenge

Plan to use Geant4 as Plan to use Geant4 as alternative simulation alternative simulation engineengine

#387 – A.Morsch, FLUKA and the Virtual Monte Carlo

Simulation toolsSimulation tools

17 February 200617 February 2006 Event Processing Applications summaryEvent Processing Applications summary 1010#257 – G.Cosmo, Recent developments & upgrades to the Geant4 geometry modeler

#371 – J.Apostolakis, Geant4 in production: status and developments


FlukaFluka

#420 – L.S.Pinsky, Update on the status of the FLUKA Monte Carlo Transport code


Parameterised SimulationParameterised Simulation

#253 – J.Weng, GFLASH – parameterised electromagnetic shower in CMS


CEDAR is primarily a generator tuning system, combining JetWeb and HepData

Aim is that any expt MC configuration should be “CEDAR-blessed” to be considered trustworthy

Modular design and many spin-offs CEDAR also hopes to encourage code

reuse and uptake of common interfaces in HEP software

CEDARCEDAR

#253 – A.Buckley, HepData & JetWeb: HEP Data Archiving and Model Validation

Simulation validationSimulation validation


Simulation validation on the GRIDSimulation validation on the GRID

#300 – A.Ribon, Geant4 Acceptance Suite for Key Observables

On candidate release 8.0 of Geant4

17 February 200617 February 2006 Event Processing Applications summaryEvent Processing Applications summary 1616#308 – M.G.Pia, Simulation for LHC radiation background: optimisation of monitoring and experimental validation

Frameworks & data modelsFrameworks & data models


Ideas from Babar, CDF, CLEO, D0, and the previous CMS system

Five types of dynamically loadable processing components• Source (provides the Event to be processed)• Output Module (stores the data from the Event)• Producer (creates new data to be placed in the Event)• Filter (decides if processing should continue for an

Event)• Analyzer (studies properties of the Event)

Components only communicate via the Event Components are configured at the start of a job

CMS Framework & data modelCMS Framework & data model

#242 – C.Jones, The new CMS Event Data Model and Framework

Ne

w


ATLAS Event ATLAS Event Data ModelData Model

Issues with Issues with schema evolutionschema evolution

Solved adopting Solved adopting an ‘Event an ‘Event Management Management Board’ and new Board’ and new tools for tools for detecting schema detecting schema changeschanges

Attempting ‘dual’ Attempting ‘dual’ EDM designEDM design

#76 – E.Moyse, The ATLAS Event Data Model

Ne

w


ALICE FrameworkALICE Framework Relatively small number of lines of codeRelatively small number of lines of code

• Fully based on ROOT, including most of their Fully based on ROOT, including most of their needed tools and specific software choicesneeded tools and specific software choices

Adopting strict code conventionsAdopting strict code conventions• No STL, templates or exceptionsNo STL, templates or exceptions

Support for many compilers/platformsSupport for many compilers/platforms No software management toolNo software management tool

• Pure CVSPure CVS

Prototyping event processing with PROOFPrototyping event processing with PROOF Data challenge: last exercise before data Data challenge: last exercise before data

takingtaking• Test of the system with simulationTest of the system with simulation• Up to 3600 jobs running in parallelUp to 3600 jobs running in parallel

Issues:Issues:• Difficulty in Planning in a widely distributed Difficulty in Planning in a widely distributed

community; difficulty in setting milestonescommunity; difficulty in setting milestones• Communication seen as a real problem in spite of Communication seen as a real problem in spite of

frequent meetingsfrequent meetings

#391 – F.Carminati, The ALICE Offline Framework


ILC frameworkILC framework RequirementsRequirements

• Easy to install, use, modify Easy to install, use, modify and extendand extend

• Must be modular and written Must be modular and written in C++in C++

• Based on the international Based on the international ILC data format LCIOILC data format LCIO

MARLIN application MARLIN application frameworkframework• Set of classes providing the Set of classes providing the

core functionalitycore functionality• Based on the concept of Based on the concept of

processorprocessor• A ‘processor task can be as A ‘processor task can be as

simple as creating one simple as creating one histogram or complex as histogram or complex as track finding and fitting in the track finding and fitting in the central trackercentral tracker

• AIDA interfaces for histo: AIDA interfaces for histo: AIDAJNI/JAIDA, RAIDA (Root AIDAJNI/JAIDA, RAIDA (Root based), …based), …

#422 – F.Gaede, A modular Reconstruction Software Framework for the ILC

ReconstructionReconstruction


RecPackRecPack C++ reconstruction toolkitC++ reconstruction toolkit Provides the common tools Provides the common tools

of any reconstruction of any reconstruction programprogram

• avoids reinventing the wheel !!!avoids reinventing the wheel !!! Modular structure allowing Modular structure allowing

for extensions in any for extensions in any directiondirection• Can plug-in Geant4 for the Can plug-in Geant4 for the

simulation step simulation step Setup independentSetup independent Being successfully used by Being successfully used by

several HEP experimentsseveral HEP experiments

#1 – A.Cervera Villanueva, RecPack, a general reconstruction toolkit


CMS CMS ReconstructionReconstruction

The code must be able to run The code must be able to run in three different scenarios:in three different scenarios:1.1. Bare root: open the POOL Bare root: open the POOL

catalog, and inspect the Data catalog, and inspect the Data Objects Objects (thus their centrality)(thus their centrality)

2.2. CMSSW-Lite: load a small CMSSW-Lite: load a small number of libraries, don’t number of libraries, don’t allow access to any allow access to any calibration, mag field map calibration, mag field map etcetc

3.3. Full-CMSSW: full access to Full-CMSSW: full access to calibrations and full calibrations and full availability of libraries.availability of libraries.

Old software (ORCA)Old software (ORCA)• Used for the Physics TDR Vol Used for the Physics TDR Vol

I, just publishedI, just published• Used by the vast majority of Used by the vast majority of

usersusers• Full functionality with the Full functionality with the

new software before Summernew software before Summer

#270 – T.Boccali, CMS Reconstruction Software


FatRas fast track FatRas fast track simulationsimulation

Fast track simulation for the Fast track simulation for the ATLAS Inner DetectorATLAS Inner Detector

Spin-off of the new extrapolation Spin-off of the new extrapolation package (simulation using the package (simulation using the new reconstruction geometry)new reconstruction geometry)

Enhanced by the modularity of Enhanced by the modularity of the new ATLAS offline track the new ATLAS offline track reconstruction, completely written reconstruction, completely written and embedded in the EDMand embedded in the EDM

Usage in validation of track Usage in validation of track finding and track/vertex fittingfinding and track/vertex fitting

Powerful tool for fast checks / Powerful tool for fast checks / strategy findingstrategy finding

#96 – A.Salzburger, The new ATLAS Fast Track Simulation engine


ALICE reconstruction ALICE reconstruction algorithmsalgorithms

Fast Hough-Transform TPC Tracking:Fast Hough-Transform TPC Tracking:• Very good efficiency (stable up to Very good efficiency (stable up to

dN/dy~8000)dN/dy~8000)• Pt resolution worsens linearly with PtPt resolution worsens linearly with Pt• ~5s comp. time for central PbPb event ~5s comp. time for central PbPb event

with dN/dy~4000with dN/dy~4000~8 Mbytes/s processing rate (compressed ~8 Mbytes/s processing rate (compressed data)data)~0.15 ~0.15 s/ADC count (hit)s/ADC count (hit)

• FPGA implementation is under FPGA implementation is under development - would allow to diminish development - would allow to diminish the computing time to hundreds of the computing time to hundreds of millisecondsmilliseconds

ITS Tracking:ITS Tracking:• Hough Transform tracks are efficiently Hough Transform tracks are efficiently

propagated to ITSpropagated to ITS• Fast and efficient ITS cluster finder, Fast and efficient ITS cluster finder,

vertex and trackingvertex and tracking• Track parameters resolution is greatly Track parameters resolution is greatly

improved (excellent impact parameter improved (excellent impact parameter resolution)resolution)

#385 – M.Ivanov, Track reconstruction algorithms for the ALICE High-Level Trigger


ATLAS track ATLAS track reconstructionreconstruction

A new framework A new framework flexible and easy to flexible and easy to maintainmaintain

Novel tracking Novel tracking technique in the MDTtechnique in the MDT• Multilayer Hough Multilayer Hough

TransformTransform Good performance Good performance

(and making use of (and making use of abstract interfaces)abstract interfaces)

#385 – D.Primor, Track reconstruction with the ATLAS detector


TPC distortions TPC distortions corrections at STARcorrections at STAR

Developed a technique to determine and apply TPC distortion corrections on an event-by-event basis

No obvious luminosity limitations

#155 – G.Van Buren, Adaptive on-the-fly calibration of TPC distorsions


Tracking in High Density Tracking in High Density Environment in ALICEEnvironment in ALICE

TRD detector is also very useful TRD detector is also very useful for reconstruction for reconstruction • was originally developed for was originally developed for

electron identificationelectron identification Local Reconstruction Local Reconstruction

• Excellent space resolution for high Excellent space resolution for high momentum track (small incident momentum track (small incident angle) angle) significant improvement significant improvement in the momentum resolutionin the momentum resolution

• Works in high density environmentWorks in high density environment The most significant improvement The most significant improvement

is due to the is due to the correct error correct error parameterizationparameterization

Local Reconstruction

#394 – M.Ivanov, Track reconstruction in high density environment


Event Selection with Event Selection with Gene Expression ProgrammingGene Expression Programming

New Evolutionary computation method New Evolutionary computation method Gene Expression ProgrammingGene Expression Programming

Applied to Event Selection Applied to Event Selection

Allows forAllows for fast identification of powerful cutsfast identification of powerful cuts signal/background separation of 92-95% signal/background separation of 92-95%

accuracy for samples with S/N = 0.25, 1, accuracy for samples with S/N = 0.25, 1, 55

potential of discovering new potential of discovering new correlations between variablescorrelations between variables

large number of selection functions large number of selection functions does not improve the classification does not improve the classification accuracyaccuracy

Still in the R&D phaseStill in the R&D phase needs software development -> needs software development ->

underwayunderway

#135 – L.Teodorescu, HEP Event Selection with Gene Expression Programming

AlignmentAlignment

17 February 200617 February 2006 Event Processing Applications summaryEvent Processing Applications summary

Track based Track based alignment in CMSalignment in CMS

HIP (Hits & Impact Points) HIP (Hits & Impact Points) iterative methoditerative method• To be used for CMS pixel To be used for CMS pixel

alignmentalignment• Basic idea: reconstruct Basic idea: reconstruct

tracks normally, then tracks normally, then align individual detectors align individual detectors or composite structuresor composite structures

• Modules or alignment Modules or alignment parameters can be free or parameters can be free or fixedfixed

• Rigid support structures Rigid support structures can be aligned as well as can be aligned as well as individual sensorsindividual sensors

• Easily parallelizable Easily parallelizable methodmethod

#356 – T.Lampen, Track based alignment of composite detector structures


Variable ATLAS Variable ATLAS detector descriptionsdetector descriptions

Basic idea: Basic idea: have possibility of have possibility of building various ATLAS geometry building various ATLAS geometry layouts with every single version layouts with every single version of ATLAS softwareof ATLAS software

ATLAS geometry versioning ATLAS geometry versioning system is based on system is based on HierarchicalHierarchical VersioningVersioning of detector description of detector description primary numbers stored in the primary numbers stored in the ATLAS ATLAS Geometry DatabaseGeometry Database

In order to switch between In order to switch between different geometry layouts it is different geometry layouts it is enough to change a single enough to change a single parameter: ATLAS parameter: ATLAS top level top level geometry taggeometry tag

ATLAS geometry tags can be ATLAS geometry tags can be passed across job boundariespassed across job boundaries• Using persistent Using persistent TagInfoTagInfo objects objects• Subsequent jobs can pick up Subsequent jobs can pick up

correct geometry configuration correct geometry configuration from the input file bypassing from the input file bypassing manual configuration through job manual configuration through job optionsoptions

#67 – V.Tsulaia, Software Solutions for a Variable ATLAS Detector Description


Silicon Tracker alignment Silicon Tracker alignment in ATLASin ATLAS

Issue: Monitor the motion of ~ 6k Si modules using Issue: Monitor the motion of ~ 6k Si modules using trackstracks

Alignment software using global fit method Alignment software using global fit method implemented implemented • with global with global 22

• Reasonable timing ~ 5ms/trackReasonable timing ~ 5ms/track First tests showed it is quantitatively possible to align a First tests showed it is quantitatively possible to align a

system with O(1000) DoFsystem with O(1000) DoF• Extension to full geometry under studyExtension to full geometry under study• More constraints to be considered, e.g. survey:More constraints to be considered, e.g. survey:

Planned: study of constraints from resonances (Z or J/Planned: study of constraints from resonances (Z or J/) ) • To extend: more statistics and iterations to recover large To extend: more statistics and iterations to recover large

misalignmentsmisalignments Coming months: align many Silicon modules to high Coming months: align many Silicon modules to high

precision - big impact on physics performanceprecision - big impact on physics performance Alternative alignment methods developed in ATLASAlternative alignment methods developed in ATLAS

• These are mainly iterative methods. No big system to These are mainly iterative methods. No big system to solve but iterations required even for small misalignmentssolve but iterations required even for small misalignments

New method under development (with Kalman filter)New method under development (with Kalman filter)Measurement errorMeasurement error

Adding mult. scattAdding mult. scatt

#84 – A.Hicheur, Global fit method for the alignment of the Silicon Tracker in ATLAS Athena


COCOACOCOA COCOA is a general purpose alignment COCOA is a general purpose alignment

software developed as a Software software developed as a Software Engineering projectEngineering project• User describes optical system in User describes optical system in

ASCII filesASCII files• COCOA reconstructs the unknown COCOA reconstructs the unknown

parameters and propagate the errorsparameters and propagate the errors COCOA stressed by years of use in CMSCOCOA stressed by years of use in CMS Full CMS Link alignment system Full CMS Link alignment system

(2865 parameters)(2865 parameters)• 25 minutes25 minutes in Athlon 1.3 GHz in Athlon 1.3 GHz• Memory: Memory: 590 Mb (590 Mb (long double matrices)long double matrices)

Time and memory scales as Time and memory scales as (#param)2 !(#param)2 !• Next challenge is to simulate full CMS (40k Next challenge is to simulate full CMS (40k

params)params)• Methods under studyMethods under study

#321 – P.Arce, COCOA: General purpose software for simulation and reconstruction of optical alignment systems

LHCb Alignment FrameworkLHCb Alignment Framework Misalignments are applied through detector Misalignments are applied through detector

structurestructure• ““Interesting” detector elements have access to misalignment Interesting” detector elements have access to misalignment

matrixmatrix• Misalignment represents change from nominal alignment Misalignment represents change from nominal alignment in in

the reference frame of the detector element the reference frame of the detector element i.e. relative i.e. relative to its parent detector elementto its parent detector element

Added runtime misalignments to detector Added runtime misalignments to detector components components

• extending the LHCb detector description frameworkextending the LHCb detector description framework

Misalignments are tied in to the Conditions Database frameworkMisalignments are tied in to the Conditions Database framework• to allow both automatic runtime updating and propagation of changes, plus to allow both automatic runtime updating and propagation of changes, plus

versioning and time dependence of alignment parametersversioning and time dependence of alignment parameters The functionality was tested within the LHCb reconstruction chainThe functionality was tested within the LHCb reconstruction chain

• LHCb subdetectors are using it to investigate detector alignment procedures and LHCb subdetectors are using it to investigate detector alignment procedures and strategies, systematic effects, etc.strategies, systematic effects, etc.

The extension is a non-intrusive enhancementThe extension is a non-intrusive enhancement• respects the design principles of the LHCb detector description suiterespects the design principles of the LHCb detector description suite

#356 – J.Palacios, The LHCb Alignment framework

Access to Non-EventAccess to Non-EventData for CMSData for CMS

• Physicists have one set of rules to learn so as to access all such dataPhysicists have one set of rules to learn so as to access all such data Design uses type-safe data accessDesign uses type-safe data access

• ‘‘on demand’ data retrieval or computationon demand’ data retrieval or computation• automatic consistency between related ‘Intervals of Validity’automatic consistency between related ‘Intervals of Validity’• C++ object lifetime managementC++ object lifetime management

Client code does not need to be recompiled or relinked when Client code does not need to be recompiled or relinked when configuration changesconfiguration changes• Implemented using dynamically loadable componentsImplemented using dynamically loadable components

Based on experience from CLEOBased on experience from CLEO• expect the system to provide for all of CMS’s future needsexpect the system to provide for all of CMS’s future needs

CMS has a unified data model for CMS has a unified data model for all non-Event dataall non-Event data• Based on an ‘Interval of Validity’ Based on an ‘Interval of Validity’

accessed through an EventSetupaccessed through an EventSetup

#241 – C.Jones, Access to Non-Event Data for CMS

VisualizationVisualizationCommissioningCommissioning


ATLAS Online ATLAS Online Histogram PresenterHistogram Presenter

OHP is an interactive application developed to present histograms produced by ATLAS Online Monitoring System

Is fully configurable (different views of the produced histograms, drawing options)

Can communicate with the histogram producers sending commands

Optimizes network traffic Manages reference histograms OHP will be used in the ATLAS

commissioning starting from March 2006

#24 – A.Dotti, OHP: An Online Histogram Presenter for the ATLAS experiment


AtlantisAtlantis Atlantis event display is a

stand-alone Java application Uses variety of 2D projections,

multiple views (windows) on canvas

A part of the ATLAS SW, depends only on Java

Uses simplified detector geometry (not a detector display)

JiveXML (written in C++) interfaces ATLAS SW framework Athena (its event store) and Atlantis

Access to the event data from Atlantis• using the event files produced

by JiveXML (offline)• reading the event data over

network from JiveXML server (online)

#33 – Z.Maxa, Event visualization for the ATLAS experiment – the technologies involved


CMS Cosmic CMS Cosmic ChallengeChallenge

The MTCC is a fundamental The MTCC is a fundamental milestone for the new CMS milestone for the new CMS softwaresoftware• Global Data Acquisition with Global Data Acquisition with

(almost) all sub-detectors(almost) all sub-detectors• Data quality monitoring, Data quality monitoring,

event selection and event selection and streaming based on local streaming based on local reconstructionreconstruction

• Non-Event Data DB Non-Event Data DB infrastructureinfrastructure

• Offline global reconstruction Offline global reconstruction using alignment datausing alignment data

CMS is on schedule to CMS is on schedule to achieve these goalsachieve these goals

#312 – G.Bruno, Software for the CMS Cosmic Challenge


CMS Data Quality CMS Data Quality MonitoringMonitoring

DQM for CMS Tracker is very challenging due its size and complexity

Tracker DQM software exists with basic functionality of the producer and consumer

A simple and very useful tracker specific visualization tool, TrackerMap exists along with a couple of generic graphical user interfaces

Important milestones in 2006:• Cosmic Challenge (CMS

slice test with Cosmics)

#283 – S.Dutta, Data Quality Monitoring for the CMS Silicon Tracker


Few considerations …Few considerations … Modular software and abstract interfacesModular software and abstract interfaces

• Experiments frameworks (CMS, ILC, …)Experiments frameworks (CMS, ILC, …)• Reconstruction softwareReconstruction software

Collaborative softwareCollaborative software• Share data and validation resultsShare data and validation results

Usage of the GRID for simulation validation and Usage of the GRID for simulation validation and regression testingregression testing

Focusing on AlignmentFocusing on Alignment• Track-based, opticalTrack-based, optical• Software is adapting, also simulationSoftware is adapting, also simulation

Looks never too late for big revolutions or major re-Looks never too late for big revolutions or major re-designsdesigns• Power of Object-Oriented ? …Power of Object-Oriented ? …


THANKS !

event processing applications summarycomputing in high energy physics 2006 – mumbai (india) event...

Documents