1 / 19

DZero Monte Carlo Production

This article provides an overview of the Monte Carlo production and verification process for the CMS experiment. It covers topics such as code verification, runtime environment distribution, physics process specification, request specification, automatic request processing, and remote processing centers.

johnsmartin
Download Presentation

DZero Monte Carlo Production

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. DZero Monte Carlo Production Ideas for CMS Greg Graham Fermilab CD/CMS 1/17/01 Hcal/Jets/Met Meeting

  2. Monte Carlo Production An Overview Greg Graham - CMS Production

  3. Monte Carlo Production Systems • Monte Carlo Verification I - Code • Patches, Versions, Executables • RED : Runtime Environment Distribution • Monte Carlo Verification II - Executables • Check representative output from all platforms • Monte Carlo Specification • Physics Process (H), Request specification (H) • Control and Tracking : The Grid • Condor pools, Globus • Special Tools Greg Graham - CMS Production

  4. Monte Carlo Verification - I • Before anything is distributed, make sure : • Correct versions of external code (eg - Pythia) are being used. • All needed patches have been applied • The output looks good at the build site Greg Graham - CMS Production

  5. Runtime Environment Distribution • Dzero : the “mini-tarfiles” • Offline production executables from official builds • Packaged with needed .so and databases • Installed on top of each other in regular way so as not to destroy previous installations • It is not optimal to distribute the build environments and compile everywhere • missing patches, different static libs, (?) compilers, etc. • sMC-site? : systematic error from site to site MC variation - minimize this, but make it measurable Greg Graham - CMS Production

  6. Monte Carlo Verification II • After executables have been distributed, make sure : • standard samples are run at different sites • output is compared from site to site Greg Graham - CMS Production

  7. Physics Process Specification Schema developed for relational DB system Keeps track of cardfiles Cardfiles required to be stored in CVS Keeps track of cardfile elements production, decay, PDF, etc. Attributes (eg Pt limit, Eta limit) and cardfile elements can be added dynamically Attributes plus cardfiles define what we mean by Physics Process This was harder than it seemed at the outset. Greg Graham - CMS Production

  8. Physics Process Specification Production Process PDF Decay Process Pt Limits Eta Limits Cvs module : Cards cardfile : mycards.pythia version : v0.1 Ttbar CTEQ5 W->jets Pythia, ver 6.134 Pt > 5.0 GeV |Eta| < 3.0 Greg Graham - CMS Production

  9. Request Specification Schema developed for Relational DB system Keeps track of processing requests List can grow dynamically; keeps track of whole processing chain Keeps track of groups of input/output files In Dzero, SAM “projects” are defined on the input/ output files Requests can be split into site-specific “work-requests” In Dzero, appearance of files with the correct parentage in SAM constitutes job tracking Greg Graham - CMS Production

  10. Automatic Request Processing Schema Greg Graham - CMS Production

  11. Monte Carlo Job Specification • mc_runjob : A D0 package written in Python • Driven by macro scripts • Macro scripts serve to specify processing steps in a regular way • Runs most D0 Offline Executables • All in principle … • Chains executables together • Run generation through ntuple in one step • Handles naming of input and output files • Handles seed generation, run numbers, etc. Greg Graham - CMS Production

  12. The Grid • Many experimental projects are underway for the benefit of CMS Production • EuroGrid, PPDG, GriPhyN • Existing Tools : Globus, Condor • Hardware : Farms & Clusters, Huge SMPs, … • The Vision : How can we go beyond production and bring the power of the hardware to individual physicists ? • Dzero : 100K events/day = 200 events/day/physicist ; CMS may certainly have more capacity. • Do we dare let anyone run their own MC like in the days of Fortran ? Greg Graham - CMS Production

  13. Specific Dzero Tools Greg Graham - CMS Production

  14. Mc_runjob “Configurators” Run and configure each D0 Offline executable The “Linker” Links Configurators Collects shell scripts that actually do the work Macro scripts The Linker and Configurators obey a regular set of macro commands in macro scripts Can easily be extended ! Greg Graham - CMS Production

  15. Inheritance Hierarchy Inheritance is used to incorporate D0 specific configuration (ie - RCP, runtime environment, naming, SAM metadata) Other special purpose configurators can handle the batch system, manage a list of files, etc. Can easily be extended ! Greg Graham - CMS Production

  16. Graphical Support While primarily a macro script driven tool, the mc_runjob inheritance hierarchy also provides core tools with GUI methods that build a GUI from information contained in the configurators themselves. No special porting task for GUI ! Greg Graham - CMS Production

  17. Automatic Request Processing System Greg Graham - CMS Production

  18. Remote Processing Centers • Monte Carlo simulation, digitization, and eventually some MC reconstruction will be handled offsite. • Currently operating RPCs • UT Arlington; IN2P3, Lyon; NIKHEF, Amsterdam; Prague • Future planned RPCs • Lancaster; Rio de Janeiro; TIFR • Generated events will be stored with SAM • We expect better than 100K events/day Greg Graham - CMS Production

  19. References • G.E. Graham “The Dzero Monte Carlo Challenge” Proceedings, CHEP 2000 • G.E. Graham “Dzero Monte Carlo” Proceedings, ACAT 2000 (to appear soon) Greg Graham - CMS Production

More Related