ETH Beowulf day January 31, Adrian Biland, Zhiling Chen, Derek Feichtinger, Christoph Grab, André Holzner, Urs Langenegger

Transcription

1 CMS SM Meeting Nov Analysis and simulation of proton-proton collision data at LHC ETH Beowulf day January 31, 2006 Adrian Biland, Zhiling Chen, Derek Feichtinger, Christoph Grab, André Holzner, Urs Langenegger

2 Overview Asgard Technical update The CMS experiment at the Large Hadron collider at CERN Physics topics Analysis overview Current Status of usage Future Plans

3 Technical update Institute of Particle Physics took over Asgard last summer Operating system upgrade: Migrated (almost all) worker nodes from RedHat 7.2 to Scientific Linux CERN 3 (SLC3) because this is the standard (and probably the only certified) platform for the LHC experiments This is basically a recompiled RedHat Enterprise Linux 3 Automatic software updates of worker and login nodes via http each night installed http proxy (squid) on login node to avoid downloading each package ~ 200 times...

4 Technical update One login server migrated to SLC3, two currently inactive (Re-)installed OpenPBS as batch queuing system, still some configuration problem with it...(and other funny behavior..) Added five new file servers (running Suse 9.3) 5 * (2 * 2.5 TB) Raid5 Two network interface cards (NICs) per server (still problems with Linux support with one of the NICs on each server)

5 Technical update Installed Ganglia for load monitoring:

6 The Large Hadron Collider Lake Geneva Geneva airport

7 The Large Hadron Collider Magnets in the tunnel

8 The CMS experiment proton beam 2 tracking chambers muon chambers crystal calorimeter proton beam 1 average physicist hadron calorimeter The Compact Muon Solenoid omni purpose experiment

9 The CMS experiment

10 The CMS experiment

11 The CMS experiment View of the underground cavern

12 The CMS experiment Detection of particles in CMS

13 Physics topics Unanswered questions in particle physics: How do fundamental particles acquire their mass? Higgs mechanism, predicts new particle Can we solve the 'hierarchy problem' (electroweak scale much smaller than gravity scale) and the 'fine tuning' problem? Do the four fundamental forces unify at a certain scale? Do we have more than three space dimensions? Do supersymmetric particles exist? Most of the theories (extensions to the current standard model of particle physics) predict new particles which we want to see in various mass spectra

14 A typical collision in the detector Hits in the tracking chamber Energy deposit in the crystal calorimeter (=photon or electron) Beam pipe Tracks of charged particles pp Higgs (+X) γγ (+X)

15 A typical collision in the detector Same event (collision), view along the beam pipe pp Higgs (+X) γγ (+X)

16 Physics topics Unfortunately, the most often occurring processes are already known and are a background for these new processes. Most famous example: pp Higgs (+X) γγ (+X) : γ = photon background process Mass of the photonphoton system signal process on top of background

17 Analysis overview collision simulation Simulated particle momenta electronics simulation Raw data (ADC counts) detector Pulse reconstruction Reconstructed pulses detector simulation Energy deposits Particle ident. Energy/momentum reconst. Reconstructed particles

18 Analysis overview Step time/event Collision simulation few 10 msec Detector simulation few minutes Electronics simulation few seconds Reconstructions minute data recorded from detector 10 msec Tasks are trivially parallelizable, collision events are (almost) independent of each other

19 Dataset sizes/requirements Data: ~109 events per year recorded on tape For signal processes typically order of 10k events Backgrounds can be order of magnitudes larger size per event: Simulation specific data files: ~ 0.5 MB Common data files: ~ 1 MB Total for simulation: ~ 1.5 MB Estimate for number of simulated events required: 0.1 ~ 5 Mio (per physics process)

20 Future plans CMS analysis activity is currently ramping up on Asgard (mostly busy now with MAGIC jobs, see previous Beowulf days presentations by A.Biland) Gain experience with running jobs for detector simulation electronics simulation reconstructions user analysis for physics processes our institute is working on

21 Future Plans Install LHC Computing Grid (LCG) software to make Asgard a Tier 3 center (Practical grid experience as a user shows that there is still a lot of room for improvement of the software involved...) Attach Asgard (Tier 3) to Swiss Tier 2 at CSCS in Manno CSCS probably will be attached to CERN or Forschungszentrum Karlsruhe (German Tier 1) (Note that a Tier 1 center is quite expensive, e.g. a tape archiving system is required, that's why Switzerland doesn't have its own)

22 Future plans

23 Future plans CMS will record its first data in Summer 2007 After startup, the main usage of Asgard will be: storing parts of reconstructed data (identified particles) and corresponding simulation developing and running user analysis programs on this data. This activity requires mostly high I/Operformance, not so much CPU time simulating additional physics processes and detector configurations with non-standard settings. Lots of CPU time required, bandwidth a secondary issue let other CMS users run on these datasets via LHC computing grid whenever resources idle