Upload
randell-black
View
218
Download
5
Embed Size (px)
Citation preview
Alice week, 10/03/2003 P.Hristov 1
Report from the off-line week
P.HristovAlice week
March 10, 2003CERN
Alice week, 10/03/2003 P.Hristov 1
Outline
■ Simulation■ Reconstruction■ NewIO■ AliEn & Analysis■ GRID
Alice week, 10/03/2003 P.Hristov 1
Simulation
■ Geometrical description: Updated geometry for ITS, RICH, FMD, START, V0, MUON, EMCAL, ABSO
■ Detector response: TPC cross-talk; MUON station 1 pad response; TRD background from n-capture in Xe
■ Event generation: jets (Pythia); gray particles; Glauber model (min. bias and hard processes)
■ Fast simulation: base classes; MUON tracking and trigger
Alice week, 10/03/2003 P.Hristov 1
Fluka: interaction and transport MC code
■ Collaboration between CERN(EP & AB), INFN, and authors signed in February 2003
■ Fluka physics:■ Hadron-nucleon: elastic, exchange,
resonance production and decay, low energy π an K interactions, Dual Parton Model + hadronisation at high energies
■ Hadron-nucleus: PEANUT (generalized internuclear cascade + preequilibrium), Glauber-Gribov model at high energies, final nucleus de-excitation
Alice week, 10/03/2003 P.Hristov 1
Fluka (cont.)
■ Neutron transport below 20 MeV: detailed multi group approach
■ Electro-magnetic interactions: photo effect, Rayleigh and Compton scattering, pair production, bremsstrahlung, Bhabha and Møller scattering, positron annihilation, μ capture, Cerenkov photons, photonuclear reactions, etc.
■ Improved charged particles transport■ Heavy ions: DPMJET at high energy; PEANUT,
relativistic molecular dynamics, Bolzman master equation at low energy
■ It will be interesting to use it!
Alice week, 10/03/2003 P.Hristov 1
VirtualMC : TFluka & Flugg
■ Simulation chain active from primary particle generation to writing of hits
■ Most of the control methods in TFluka implemented
■ Most of the dynamic track/particle properties available in TFluka
■ Volume/material/medium related functionality delegated to Flugg
■ Work in progress: remaining dynamic properties, particle history, active stepping actions, physics configuration, external decayer, etc.
Alice week, 10/03/2003 P.Hristov 1
VirtualMC: Geant4■ Updated examples E01 and E02, new example
E03 prepared■ New class TMCVerbose to have the same output
from Geant3 and Geant4■ Performance tests using E03: G3/G4~0.6,
G4VMC/G4native~1.2, slow simulation in native G4 when vacuum+field
■ New users of VMC: OPERA, CBM => new requirements (documentation, etc.)
■ Converter between Root Tgeo geometry description and Geant4 developed to facilitate the comparison
■ Geant4 5.0: improved low energy hadron physics
Alice week, 10/03/2003 P.Hristov 1
Geometrical modeler
■ Navigation functionality:■ “Where am I?”, “Distance to next boundary”,
“Safety”: done and working. ■ Normals to crossed surfaces: ongoing
■ Validation of the modeler: using the database of existing G3 detector description:■ “Where am I?” is significantly faster in all cases■ “Distance to next” can be optimized in case of
very simple geometries, in general it is faster than G3
■ The precision of the modeler is much better■ Improved visualization: perspective view, focusing
Alice week, 10/03/2003 P.Hristov 1
Geometrical modeler■ Geometry checker: detection for illegal
extrusions and overlaps■ Good efficiency (detects ~90% of problems)■ Very fast (20s for full Alice geometry)■ Estimates the size of the extrusion/overlap■ Alice: ~160 overlaps/extrusions bigger than
1mm. Some of them corrected during the week, the work has to be continued!
■ TGeo integration with VirtualMC:■ Geant3: New TGeant3 for G3 standalone,
TGeo standalone, TGeo+G3 (for comparison)■ Fluka (Flugg replacement by TGeo): in
progress
Alice week, 10/03/2003 P.Hristov 1
Reconstruction I
■ TPC parallel tracking: ■ Cluster finder: threshold effects corrected
using extrapolation to Gaussian shape => cluster RMS estimation
■ In case of big RMS: Cluster unfolding using fast spline approximation
■ Track seeding with vertex constrain (2 iterations)
■ Track seeding without vertex constrain: Based on groups of nearest clusters from different pad rows, prolongated using polinomial fit (several iterations at different radii)
Alice week, 10/03/2003 P.Hristov 1
■ Parallel Kalman tracking: allow cluster sharing between tracks. Consists of prolongation, error estimation, track parameter's update.
■ Removing of overlapped track candidates (at 3 different stages)
■ dE/dx evaluation: truncated mean (60%), no shared clusters, signals at maximum
■ Results: Improved efficiency ~97% for primary particles (~90% before); Improved PT resolution by 20%; Improved dE/dx resolution ~6.8% (8.7% before); Better angular resolution by ~20%.
Reconstruction II
Alice week, 10/03/2003 P.Hristov 1
Reconstruction III■ Reconstruction of kinks and V0's:
Combinatorial search with following criteria■ DCA■ fiducial volume■ cluster density before and after the kink/V0■ Very preliminary results: 229 kinks found
out of 1000, 12 fake ones■ ITS tracking V2:
■ Adapted to the new TPC tracking (V3): ~10% gain in efficiency (now at 80% level)
■ Many other ideas tested
Alice week, 10/03/2003 P.Hristov 1
Reconstruction IV■ TRD tracking:
■ Energy loss of electrons due to bremsstrahlung included in track propagation => improved resolution for high momenta (~3%)
■ Tracking with tilted pads: back-propagation from TPC. Strange Pt dependence of the momentum resolution to be investigated.
■ Reconstruction of track length and evaluation of time-of-flight hypothesis: Propagation of track parameters TPC->ITS->Vertex(start time)->ITS->TPC->TRD■ Time resolution: depends on the momentum
resolution for slow particles (protons)
Alice week, 10/03/2003 P.Hristov 1
ESD class: discussion
■ ESD can be used to store the results of the combined tracking. It will used as a seed for track propagation between different detectors. The ITS, TPC, and TRD tracks will be kept as before.
■ Parallel ESD structure filled with MC information can be used for comparison and resolution studies.
■ Back-propagation: A possibility for base Propagator class can be investigated
■ The implementation has to start as soon as possible!
Alice week, 10/03/2003 P.Hristov 1
NewIO
■ Goal: Standardized I/O, decoupled from AliRun and AliDetector; Systematic usage of different files for data storage; Data exchange trough folders
■ Design: Based on Loaders■ Loader's responsibility: opening/closing files;
posting/accessing data■ Run Loader: works with central data (AliRun
object, event header, kinematics, stack, track references, AliRunLoader object itself)
■ Detector loader: works with detector specific data (hits, sdigits, digits, rec. points, tracks)
Alice week, 10/03/2003 P.Hristov 1
NewIO II
■ Architecture■ AliRunLoader has a list of AliLoaders■ Each detector registers its Loader to the list■ Each detectors keeps pointer to its Loader
for fast access (but does not own the Loader!). This pointer is restored automatically if one reads the detector from a file.
■ Each data type (hits, digits, etc.) is defined completely by instance of AliLoaderDataInfo class
Alice week, 10/03/2003 P.Hristov 1
NewIO III■ Status
■ Merged with v3-09-04■ ITS, TPC, and PHOS chains work correctly:
(s)digits->clusters->track(let)s->PID■ Internal redesign in progress (+AliMC, AliRec,
etc.)■ Advantages:
■ I/O decoupled from AliRun■ High stability and robustness■ Easy to use and extend■ Possible co-existence of different data versions
■ Decision: Move to NewIO on April, 28!
Alice week, 10/03/2003 P.Hristov 1
AliEn
■ 32 sites (5 with mass storage); 9 virtual organizations
■ 12 production rounds, 22773 jobs, 2428 failed (10%) => 20 Tb of simulated data produced in various places
■ Up to 450 concurrent jobs managed by 0.5 operators
■ Active development:■ AliEn file system for Linux■ Support for analysis/PROOF interface■ EDG interface
Alice week, 10/03/2003 P.Hristov 1
AliEn II
■ Internal DB API &optimization
■ Packman & Port to Itanuim/Sun
■ Web portal redesign
■ GRID partitioning■ Queue optimization■ Service
registration/discovery using P2P technology
■ Redundant hardware platform setup and XML message routing
■ Java interface■ Automatic
documentation■ UML model■ Certificate based
authentication for LDAP
■ ...
Alice week, 10/03/2003 P.Hristov 1
AliEn & Analysis■ Goal: AliEn interface for ROOT to provide an off-
line parallel analysis facility■ AliEn Generic File Access Service: allows
partial file read/write access via sshd (or rootd) daemon
■ TalienFile provides alien:// plugin for ROOT■ Job Control Interface (TAlienJobIO) for
submission, job status query, and job splitting■ TAlienAnalysis class for submission of jobs.
Interacts with the AliEn catalog to register the output files (MC type job), or to query for specific input data (Analysis/Filtering type job)
■ Almost ready for user's analysis!
Alice week, 10/03/2003 P.Hristov 1
PROOF
■ PROOF and Data Grids■ Many services are a good fit: Authentication,
File Catalog, Replication services, Resource Brokers
■ Phased integration: Static configuration, Use of one or multiple Grid services, Driven by Grid infrastructure
■ Package manager now available: distributes user environment to PROOF sandbox (source, shared libs, macros, etc.)
■ Scalability tests: good results■ Release after CHEP!
Alice week, 10/03/2003 P.Hristov 1
GRID■ EDG testbed:
■ Monitoring: one short job per RB per CE per day. Computed/scheduled~98%
■ Real production: simulate and reconstruct 5000 central Pb-Pb events; analyse these events. Started last week...
■ Test AliEn-EDG inteface: use EDG as one huge AliEn site
■ AliEn-EG interface:■ Interface site is an EGD User Interface
machine which runs the AliEn client site: Cluster monitor, CE and SE
Alice week, 10/03/2003 P.Hristov 1
GRID II
■ Main functionality: job submission, data catalog registration, and data access
■ Open issues: error and problem tracking, automatic conversion of InputData requirements, possible bottlenecks for cross-world jobs, testing
■ More news to come!
Alice week, 10/03/2003 P.Hristov 1
LCG Application area developments
■ ALICE & ROOT original proposal was to base LCG software on ROOT
■ This has been refused by the other experiments and several projects have been launched
■ POOL for persistency, initially (?) based on ROOT I/O
■ SEAL for base library■ PI for physicist interface■ Simulation project
■ Strong interest in Virtual MonteCarlo and FLUKA
Alice week, 10/03/2003 P.Hristov 1
LCG Application area developments
■ Not a very satisfactory situation■ We are heavily marginalised once more■ A lot of manpower is wasted trying to rewrite ROOT■ A situation of tension that we already knew at the
time of LCB is creating■ Few (important) things on the positive side
■ Support for ROOT■ A hope to see manpower assigned to experiments■ Interest in the Virtual MonteCarlo and in FLUKA
Alice week, 10/03/2003 P.Hristov 1
Alice off-line development
■ ALICE Offline is quite different from the other offline projects
■ Centrally managed -- no diverging lines■ A large part of the CORE is done at CERN
■ However■ Very little help from the outside for CORE business■ Sometimes CERN had to “fill-in” for some important tasks
■ The model has worked till now, but can this continue?■ CERN manpower is becoming more and more limited and
hard to get■ It is perhaps time to see whether the development can
be made more distributed■ NewIO is suffering a lot from this lack of feedback
Alice week, 10/03/2003 P.Hristov 1
Plans
■ For the moment we do not have a detailed long term planning
■ The more important “appointments” are■ Computing data challenges every year with DAQ■ Physics data challenges 1H04, 1H06
■ The bulk of the work this year will be■ Prepare for PDC 1H04 and of also to “define” it
■ What do we want to do?■ What will be the involvement of people outside
■ Prepare for ADC 5 (1H03)■ Again what will be the involvment of people
outside?