ALPGEN MC Generation at SPRACE

Main Goals

  • Make ALPGEN work out of the box for large-scale production.
  • Deliver whatever the production team needs in order to produce ALPGEN samples centrally.
  • Make SPRACE a center of ALPGEN production and distribution to the T2's in South and North America.
  • Deadline: May 2009.

Tasks and Milestones

Task Deadline Status
Working team at SPRACE Dec/2008 COMPLETE5
Interface ALPGEN-CMSSW Jan/2009 COMPLETE5
Software requirements Jan/2009 COMPLETE2
User documentation Feb/2009 COMPLETE3
Usage and tuning of ALPGEN Feb/2009 COMPLETE2
ALPHA algorithm in ALPGEN Mar/2009 COMPLETE0
ALPGEN modification May/2009 COMPLETE0

Tasks Legend

Working team at SPRACE

  • Marco Dias: software installation and support, both for large-scale production and standard users. Also understanding of inner workings of ALPGEN.
  • Thiago Tomei: ALPGEN results validation. ALPGEN integration in CMSSW. User training.
  • Flavia Dias: ALPGEN results validation.

Interface ALPGEN-CMSSW

  • The usage of ALPGEN within CMSSW is mature. The main entry point to the documentation is this page.
  • The package AlpgenInterface is the software that interfaces the output of ALPGEN (.unw files) to CMSSW.
  • Tested and validated for CMSSW_2_X. Ready for large scale production.
    • Tag V00-03-00-06 is suitable for large scale production with CMSSW_2_2_9
  • The AlpgenInterface has also gone through heavy development to adapt itself to the CMSSW_3_X Generator restructuring proposal, detailed here.
    • The responsible for this is Thiago. The dedicated Twiki is AlpgenRestructuring.
    • UPDATE Migration completed. Tag V01-01-01 is suitable for large scale production with CMSSW_3_1_X
  • Future developments
    • New project: ALPGEN --> MCDB --> CMSSW workflow

Software requirements

  • Identification of resources and tools.
    • ALPGEN v213 installed centrally at: $CMS_PATH/slc4_ia32_gcc345/external/alpgen - checking for availability, usability, possibility to upgrade.
      • Available and usable. Upgrade is difficult, depends on centralized decisions and deployment.
    • Grid / CRAB support to run ALPGEN jobs, i.e. to run production of weighted, unweighted and matched events in one go. Steps have been identified by Marco.
      • Installation and configuration of ProdAgent. Done by Marco. To be tested.
      • Installation of a local DBS for the locally stored production
      • Understanding and generation of the Workflow.xml files
      • Understanding how to modify grid2.file and/or cmsGen in order to optimize/include New Physics there
    • generators081215.pdf: A presentation describing large-scale Monte Carlo productions in CMS.

User documentation

  • Tutorials and user support have to be present.
  • Documentation must be well-organized.
  • Extensive use of dynamic pages (Twikis).
  • Tutorials included in AlpgenInterface package.
  • Doxygen documentation.

Usage and tuning of ALPGEN to generate MC events

  • Usage of ALPGEN: understood. More work is needed in usage of ALPGEN in the "unmatched" (ickkw = 0) mode.
  • ALPGEN tunings both to data (Tevatron / LHC) and to other generators is desired, in order to have handles cross-check / cross-validate the Monte Carlo productions.
  • Validation of results at different levels
    • Software level (debugging)
    • Physics (generation) - HepMC level
    • Physics (reconstruction) - RECO level
  • Development of standard, consistent validation suite.

ALPHA algorithm in ALPGEN

  • The ALPHA algorithm is the heart of ALPGEN. It must be studied thoroughly in order to understand the inner workings of ALPGEN. This knowledge will allow us to modify ALPGEN in order to better suits.

ALPGEN modification

  • Work side-by-side with ALPGEN authors.
  • Inclusion of New Physics.
  • Dependent on all other tasks: must understand how it works and how to use it, before tinkering with it.

Useful Links

References

Feedback

11/25/2008
  • Marco becomes expert like Thiago on the alpgen physics/software/running; Thiago is the official responsible at the generators group for ALPGEN (with oversight from Maurizio and myself and Roberto Ciericci)
  • Marco with the help of Guillelmo-Gomez Ceballos from MIT, and Massimiliano Chiorboli from Catania (he knows how to run massive physics production in T2, I had him do this for trigger validation reasons in Catania and Bari) and Daniele Spiga (PhD student/CERN/Padova the expert on CRAB matters) will do some R+D on how to best produce these complicated MCs; (ProdAgent, CRAB, cmsdrive whatever) directly on a Tier2 and then exercise some workflows: move them to T1 for backup (and publication?) , publish them to two other Tier2s .
  • Marco with Thiago will try to see how to put new physics (EDs and other composite models) in ALPGEN (with oversight from Maria Spiropulu) with a few simple examples that will be relevant to the ZZ and WW resonances (that Thiago/Flavia) are working.
  • Objective: have ALPGEN working out of the box for production and that we deliver whatever the production team wants to them to produce these also centrally by ~May 2009.Sao Paolo T2 can become a center of ALPGEN production for CMS needs and distribution to the rest of the T2s in South and North America.
  • Flavia to be as much as possible involved in the ALPGEN physics interface, etc (Thiago for training her and teaching her everything he knows)

Physics Motivation: According a discussion with Gino Isidori, he summarized the phenomenological situation (about the WW and WZ analyses) as it follows: Even if it might be that a Z'-like (inclusing RS graviton) resonance can decay to ttbar, the S constraint from EWPO implies that it has to be fermiofobic. At the contrary, even if ttbar is an allowed channel, WW and WZ are required in order to stabilize the EW physics. So, according to him (he wrote a first paper on this, he is writing a second in these days) WW is a must for any model with a strongly coupled new resonance. His conclusions, he said, are based on very general statements, not on specific model assumptions. He suggested to take care of the raising function in WW scattering. So, most probably, Thiago will need some assumption on the Higgs mass to control the shape of the SM bkg in the dijet invariant mass

The initial paper is here

  • Marco will start looking whether this is possible these new higgless models with dibosons to be implemented in ALPGEN. Marco will start looking at the SM diboson implementation in ALPGEN and see where to kludge the new couplings there and also put the RS graviton from Pythia to Alpgen (i.e. instead of pp->WW , pp->RS_G->WW or pp->H->WW.)
12/01/2008

Talking with Gillelmo : The way that ALPGEN is used in ProdAgent is as follows: cmsGen (a simple python script) can run AlpGen or MadGraph (the first one faster when dealing with a great number of jets). It consumes a configuration file grid2.file and generates Weighted and Unweighted events. They usually throw away Weighted events, using Unweighted as input for cmsRun.

There is another way, cmsRun can read this data directly from MCDB (ascii files), but the steps to construct this ascii files are described here.

The way to run in a Tier-2 a private Monte-Carlo production is 1)using crab or 2)using ProdAgent. In the first one we have to specify manually all the steps. In the last one you have only to generate the workflow, and running it should be so simple as running input Workflow.xml. This is the way chosen by the Higgs groups, like Aachen. The difficult part is installing and configuring ProdAgent properly. The steps should be:

  • Install and configure ProdAgent properly in our farm;
  • Install a local DBS for the local stored production;
  • understand how to generate the Workflow.xml;
  • understand how to modify grid2.file and/or cmsGen in order to optimize/include new Physics there.

After that (and the upgrade of our storage) we can seriously think in joining Exotica group.

-- ThiagoTomei - 12 May 2009

Topic revision: r3 - 2009-05-14 - ThiagoTomei
 

This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2024 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback

antalya escort bursa escort eskisehir escort istanbul escort izmir escort