1. Vincenzo Innocente, CERN/PH Core Application Software 30 November 2004, Minsk CMS RDMS Conference CMS Core Software for Physics Applications Vincenzo Innocente CERN/PH/SFT Software services for Simulation, High Level Trigger, Reconstruction and Analysis

2. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software CMS (offline) Software Slow Control Online Monitoring Distributed Persistent Data Store Manager Environmental data store Request part of event Simulation G3 and or G4 store Data Quality Calibrations Group Analysis User Analysis on demand Request part of event Request part of event Store rec-Obj and calibrations Quasi-online Reconstruction Request part of event Store rec-Obj Event Filter HLT

3. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Requirements (from the CTP, dec.`96)  Multiple Environments: u Various software modules must be able to run in a variety of environments from level 3 triggering, to individual analysis  Migration between environments: u Physics modules should move easily from one environment to another (from individual analysis to level 3 triggering)  Migration to new technologies: u Should not affect physics software module

4. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Requirements (from the CTP)  Dispersed code development: u The software will be developed by organizationally and geographically dispersed groups of part-time non-professional programmers  Flexibility: u Not all software requirements will be fully known in advance Not only performance Also modularity, flexibility, maintainability, quality assurance and documentation.

5. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Federation wizards Detector/Event Display Data Browser Analysis job wizards Generic analysis Tools ORCA FAMOS LCGtools GRID OSCAR COBRA Distributed Data Store & Computing Infrastructure CMStools Architecture Overview Consistent User Interface Coherent set of basic tools and mechanisms Software development and installation IGUANA

6. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software (O/R) DBMS (O/R) DBMS GEANT4 HEP Foundation Libraries HEP Foundation Libraries Interactive Analysis Tool Interactive Analysis Tool C++ Standard Library + Extension Toolkits C++ Standard Library + Extension Toolkits Component Architecture Framework Layering Calibration Objects Calibration Objects Generic Application Framework Physics modules Grid-Uploadable BasicServices Adapters and Extensions Configuration Objects Configuration Objects Event Objects Event Objects (Grid-aware) Data-Products SpecificFrameworks Event Filter Reconstruction Algorithms Physics Analysis Data Monitoring LCG

7. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Software Development Process Baseline u Contribute to the design and implementation of common software as part of the LCG project u Adopt common component and integrate them into our framework and applications u Milestones (timescale and features) must be negotiated with LCG and the other LHC experiments u Large use-case coverage, great flexibility, more complexity, some overhead Fallback u Do it ourselves u Full control and ownership of design, implementation and maintenance u Complete freedom in defining milestones u Streamlined implementations of limited use-cases relevant to us NOW

8. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Contribution to LCG Application Area SPI : u OVAL u SCRAM SEAL: u Contribution to general design u Implementation of Foundation classes (re-engineering of iguana classlib) u Implementation of core framework services (plug-in mechanism) u Mathlib POOL: u Responsibility of FileCatalog u Implementation of Collections based on root-tree u Contribution in the Relational storage service u Debug and testing PI: u Project Leadership u Interface to POOL Simulation: u Leadership of and contribution to the Generator task u Contribution to Simulation validation 3D (Distributed DataBase Deployment) : u Responsibility in the Proxy system (Frontier project)

9. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Simulation, Reconstruction, Analysis Software Structure Iguana Core Visualization GUI and Scripting Services COBRA Core Framework(s) Core Services IguanaCMS Detector&Event Visualization FAMOS Fast Simulation Mantis G4 Simulation Framework DDD Detector Description Framework CARF Reconstruction Framework Profound PRS Foundation Persistency Layer Application Infrastructure Generator Interface G3 OO Interface OSCAR Detector Simulation ORCA Detector Reconstruction HLT Physics Analysis

10. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Core Software componentscomponents Core Software componentscomponents  Foundation  Persistency  Event processing Framework  Book-keeping, Configuration & Conditions  In collaboration with Data Management  Frameworks and tools specific for  Simulation  In collaboration with SPROM  Reconstruction  In collaboration with RPROM  Analysis  In collaboration with APROM  Visualization  In Collaboration with APROM

11. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software OSCAR overview Object Oriented Simulation for CMS Analysis and Reconstruction Full CMS simulation based on the Geant4 toolkit Geant4: physics processes describing in detail electro-magnetic and hadronic interactions; tools for the CMS detector geometry implementation; interfaces for tuning and monitoring particle tracking CMS framework: application control, persistency, common services and tools (magnetic field, generator interfaces and support for MC truth, infrastructure for hits and readout units,…), “action on demand” to selectively load desired modules, configure, tune application CMS changed from CMSIM/GEANT3 to OSCAR/GEANT4 end 2003; OSCAR used for substantial fraction of DC04 production; will be used for physics TDR production CPU: OSCAR  1.5 x CMSIM - with lower production cuts! Memory: ~110 Mb/evt for pp in OSCAR  100 Mb in CMSIM Robustness: ~1/10000 crashes in pp events (mostly in hadronic physics) in DC04 production to 0 crashes in latest stress test (800K single particles, 300K full QCD events)

12. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software OSCAR Interfaces and services Application steering handled by CMS framework; CMS RunManager implements functionality required for G4 running and provides handles to the G4 run, event, track and step - as required for application configuration and monitoring; manages random number and cross-section table storage and retrieval Detector geometry construction automated via Detector Description Database which converts input from XML files managed by Geometry project; XML files selected by user-defined configuration Generator input (via RawHepEvent CMS format and recently HepMC) converted to G4Event; specific generator type and event format (particle gun, Pythia, etc from ntuple, ASCII, database etc) run-time configurable Interface from CMS magnetic field services to G4; field selection run-time configurable; propagation parameters via XML Infrastructure for physics lists (run-time selection of list and process types, optional activation of  /e-nuclear and synchrotron radiation,misc. customizations) and production cuts (the latter via XML) User actions (monitoring, tuning) via dispatcher-observer pattern for pointer to observable entity Persistency, histogramming, monitoring etc transparently through CMS framework (COBRA)

13. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Parameterized Simulations Implementation of fast EM shower simulation in Geant4/OSCAR, using GFLASH parameterized showers (spot density) - tuning in progress… Timing studies Geant4 6.2 - full vs fast G4FLASH 50 GeV electrons

14. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Heavy Ion Simulation G3/CMSIM: chop event in slices of 100 tracks each and run them separately; needed due to limitations from ZEBRA OSCAR/Geant4: run full HI events Factor 5 performance improvement by improved calorimeter track selection and hit processing Event cut in slices of 100 particles Full event CMSIM - G3230 min OSCAR_2_4_5 – G4 5.2320 min OSCAR_3_4_0 – G4 6.2180 min CMSIM - G3Not possible OSCAR_2_4_5 – G4 5.21010 min OSCAR_3_4_0 – G4 6.2210 min (*) …effect entirely negligible in pp events! time/evt in given machine (*) 2.3 CPU hrs on P4 3.2 GHz performance optimization with a twist… 55K generated particles, with 97K tracks from 80K vertices kept at the end of event

15. 30 November 2004, Minsk CMS RDMS conference IGUANA and IGUANACMS IGUANA and IGUANACMS are two different SCRAM-based projects. IGUANA defines a generic object model and a framework for interactive 2D and 3D visualisation. It provides a number of services/tools to generate and manipulate those objects and to manage user interactions. u iguana command, iguana studio IGUANACMS is a composite framework linking IGUANA and other frameworks such as COBRA and Geant4. It provides modules and services that shape the abstract object model of IGUANA into a concrete CMS-specific application. u iguana studio session setup

16. 30 November 2004, Minsk CMS RDMS conference IGUANA Applications and Frameworks Uniform user view of the CMS visualisation applications for ORCA, OSCAR, COBRA, Geant4, Mantis, DDD, and test-beams. Non-intrusive, loosely-connected frameworks Coherent user view Integration Uniform transparent data access Run-time environment Distribution Large-scale reuse of object-oriented libraries requires frameworks The framework provides a context for the components in the library to be reused Wrapped loosely coupled components

17. 30 November 2004, Minsk CMS RDMS conference IGUANACMS Today Visualisation applications for ORCA, OSCAR, test-beams (DAQ application); Visualisation of reconstructed and simulated objects: tracks, hits, digis, vertices, etc.; Full DDD detector visualisation; Magnetic field visualisation; Interactive modification of configurables at run-time; Custom tracker selection; Event browser;

18. 30 November 2004, Minsk CMS RDMS conference DST Visualisation List of containers in the event: updated for each event. Name and version for RecCollection. RecMuon TTrack Formatted text information for selected RecCollection iCobra event browser: graphical structure of event

19. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software CMS Data Event DataSet DataSetBookKeeping Event Electrons Electrons Tracker Alignment Tracks Tracks Ecal calibration Ecal calibration User Tag (N-tuple) Event-Collection Meta-Data Environmental data u Detector and Accelerator status u Calibrations, Alignments (luminosity, selection criteria, …) … Event Data, User Data Navigation is essential for an effective physics analysis Complexity requires coherent access mechanisms Event Collection CollectionBookKeeping

20. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Hybrid Store CMS vision of an Hybrid Store is of a Data Management System whose components do not come all from a single “vendor” u Use best technology for each “domain” è ROOT trees for event data è XML for simple descriptions and as light exchange format è DBMS for robust management of configuration and conditions è “Grid” technologies to distribute data u It is User/CMS/LCG/HEP responsibility to glue all together Challenge: produce a Data Management System u POOL as access layer è Coherent Application Interface è Flexible data “transfer” protocols (rfio/xrootd/frontier) u Common CMS Book-Keeping system u Support of native tools for experts/browse/debug

21. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Technical items on Critical Path Access to RDBMS from physics-applications u Required to access Conditions/calibrations, configuration and Metadata u Proposal is to follow Frontier model Interactive access to ORCA objects u Root is currently the most popular interactive analysis tool in HEP è We must provide easy access to ORCA objects from the Root prompt u Two solutions (could coexist) è ORCA writes a data structure that Root can understand in native mode ORCA writes in root-tree using “simple” data structures Useful for simple data inspection è We develop a root-plugin to use ORCA from Root (or Root in ORCA) LCG/AA provides plugin for pool and mechanisms to populate CINT dictionary We provide a root-binding from transient COBRA/ORCA classes Useful for analysis of “medium” complexity u Requires changes to CMS Event Model

22. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Database types We will have 4 kinds of data classes in CMS: u Construction data è All information about the sub detector construction up to the start of integration u Equipment management data è Mainly hardware description and location history u Configuration data è All information required to bring the detector in any running mode u Conditions data è All information about detector conditions

23. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Construction data Sub detector specific Sub detector responsibility Already existing for most sub detectors. Have to be available for the lifetime of CMS Data: construction info, initial calibration data, work flow info, etc. Data will partially be copied into the 3 other data classes and then frozen in the construction DBs.

24. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Equipment management data Common for all CMS Set up by CMS integration, data entered and maintained by sub detectors Data: detector geometry and electronics parts, cables, racks, crates, location history of all items, etc… User interface existing (rack wizard) to enter required information. Needs time stamping.

25. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Configuration data Sub detector specific parts on front end electronics configuration, combined DAQ, Trigger and DCS info. Set up by corresponding groups Interface to online software framework XDAQ existing First prototype of interface to DCS existing Needs time stamping, versioning and tagging

26. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software The conditions data All parameters describing the run conditions and logging. Most of the conditions data stays at the experiment and is NOT used for offline reconstruction Set up by corresponding groups Conditions data is the logging of the online system Conditions data will be used for error tracking in the online system Conditions data will be related to u Other conditions data u Event quality monitoring u Configuration data u …

27. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Data flow Configuration Conditions Calibration O ffline R econstruction C onditions DB ON line subset Equipment Configuration Conditions Create rec conDB data set Offline Reconstruction Conditions DB OFfline subset Master copy Bat 513 O nline M aster D ata S torage

28. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software DataBase Work packages OMDS u Data maintenance tools u Data visualization tools u “Export” to ORCON ORCON u Transfer to/from ORCOF ORCOF u Data distribution to/from TierX è ORACLE replication è FroNtier

29. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software FroNtier Overview CDF Persistent Object Templates (Java) FroNtier components in yellow Client Caching FroNtier Server Database FroNtier Client API Library Squid Proxy/Caching Server FroNtier Servlet running under Tomcat Database (or other persistency service) XML Server Descriptors DDL for Table Descriptions C++ Header and Stubs JDBC HTTP

30. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software A Multi-tier Architecture Database (Oracle, MySQL, etc.) Database Access Layer u Tomcat: servlet management engine u JDBC: Database connection u Database Connection pool management Caching and proxy layer: Squid u widely used, highly configurable, caching proxy server Client u Client API provides simple interface packaged in a compact client library

31. General Deployment Overview Dedicated Squid Node Tomcat Node 1 Tomcat Node N Tomcat Node 2 CAF 1 Processor Farm Database Dedicated Squid Node CAF N Processor Farm Dedicated Squid Node CAF 2 Processor Farm Squid Node N Squid Node 2 Cache Layer local to each CAF Gbit ethernit Intermediate Cache Layer DB Access Layer Squid Node 1 Load Balancing and Failover Launchpad WAN

32. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Schedule Nov/Dec 2004 u Release production software for ORCA (dst production) u Review Data Model (Event, Conditions, Book-Keeping) and propose a new one Jan/Feb 2005 u Prototype&test Conditions Database u Prototype Event Data Model and DataSet Book-Keeping system March/April 2005 u Release Analysis software u First version of new data model u First examples of full chain analysis for the Integration test July 2005 u Deploy first version of software for integration test for ALL detectors Oct 2005 u Start Integration (Magnet) test

33. 30 November 2004, Minsk CMS RDMS conference Vincenzo Innocente, CERN/PH Core Application Software Summary CMS has developed and deployed a Core Software System based on LCG components u Supports Simulation, Reconstruction, Analysis and Visualization u Successfully used in production and user analysis A process to review our Data Model & Framework in light of successful experiences at Tevatron and to account for the new Computing Model is underway u Proposal in Dec, release in April Mayor technical challenges in near future u Database access and deployment u Data access from root u Dataset book-keeping Next year milestones driven by u TDR preparation u Integration (magnet) tests