Introduction to TAGs Elizabeth Gallas Oxford ATLAS-UK ATLAS UK Distributed Computing Tutorial Edinburgh United Kingdom Edinburgh, 29-30 January 2009 30 Jan 2009 Elizabeth Gallas 1 Outline Introduction: TAG feedback f db k (meet ( the h developers) d l ) What is a TAG ? TAGs in the ATLAS Computing model TAG Contents and Formats ELSSI – Event Level Selection Service Interface A web interface using TAGs for ATLAS event selection Important TAG related hyperlinks Hands on tutorial S Selecting l ti events t off interest, i t t refining fi i selection l ti criteria it i Extracting that selection Using selection as input to Athena analysis 30 Jan 2009 Elizabeth Gallas 2 The TAG Development p Group p • Argonne: David Malon, Jack Cranshaw, Qizhi Zhang, Peter van Gemmeren • Glasgow: MikeKenyon, HelenMcGlone, TomDoherty • CERN: Florbela Viegas, g , Trivan Pal • Oxford: Elizabeth Gallas • Chicago: Marco Mambelli • Grenoble: Solveig Albrand • Orsay y LAL: Julius Hrivnac • Sheffield: Tulay Donszelmann (PAT) Users send feedback to: ATLAS Physics y Metadata Hypernews yp 30 Jan 2009 Elizabeth Gallas 3 What’s a “TAG” ((in this talk)) ATLAS “TAG” overload: Not Not Not Not Not Not an acronym (retrospectively TAGs Are Great !) a CVS tag or a release l tag what is tracked by the ATLAS Tag Collector a COOL (Conditions Database) Tag a configTag or a project tag a B, flavor, or muon tag ATLAS TAGs are: In this talk: About 200 indexed variables (Event-level metadata) for each event Size: ~1 kB/event -> 1 TB/year In TAG files – distributed dist ib ted to all Tie Tier 1 sites In a relational database (Oracle) – distributed some Tier 1 In 2 format: ATLAS TAG application is A system using ATLAS TAGs A variety of supporting tools and infrastructure PURPOSE: Facilitates event selection for analysis 30 Jan 2009 Elizabeth Gallas 4 ATLAS TAGs in the ATLAS Computing p g model Stages of ATLAS reconstruction ¾ RAW data file ¾ ESD (Event Summary Data) ~ 500 kB/event ¾ AOD (Analysis Object Data) ~ 100 kB/event ¾ TAG (not an acronym) ~ 1 kB/event TAGs Are produced in reconstruction Can C b be re-produced d d iin re-processing i Customized TAGs can be defined/produced TAG formats: File based (produced in reconstruction chain above): AthenaAwareNTuple format (AANT) Oracle Database (populated in ‘upload’ upload process): Available globally through network connection Adding ‘Run Metadata’ at Temporal, Fill, Run, LB levels Used by y the TAG Browser (ELSSI) ( ) 30 Jan 2009 Elizabeth Gallas 5 What does the “TAG” contain ? “TAG content”: Your event selection variables ~200 200 variables/event: Event identification (run, event, lumi block, time …) Trigger decisions at all three levels (bit encoded) Numbers of electrons, muons, photons, taus, jets pT, eta, phi for highest-pT objects Global quantities (e.g., scalar and missing ET) Detector status and quality words Physics and performance group status words references to AOD,ESD,RAW for back-navigation PAT (Physics Analysis Tools) PAT group supports t the th ‘Official’ ‘Offi i l’ TAG content t t with ith input from the physics groups (YOU !) See the PAT TWiki: TagForEventSelection g 30 Jan 2009 Elizabeth Gallas 6 TAG related hyperlinks: yp TWiki with TAG content in ATLAS release: https://twiki https://twiki.cern.ch/twiki/bin/view/AtlasProtected/ cern ch/twiki/bin/view/AtlasProtected/ TagForEventSelection TWiki for Tutorial: https://twiki.cern.ch/twiki/bin/view/Atlas/EventTag Tutorials 4 Tutorial T t i lC Components t ELSSI (Browser based) File File-based based TAGs (Command line based) JiveXML with TAGs (Atlantis Event Display) TAG based selection with PAthena on the Grid ELSSI - Event Level Selection Service Interface: https://voatlas18.cern.ch/tagservices/index.htm 30 Jan 2009 Elizabeth Gallas 7 ELSSI: Implementation p & Features ELSSI is a web based supporting tool (using TAG Database) Running g on server at CERN ((PHP and Javascript p on backend) Supported: Firefox, Safari (not IE) H Helps l you navigate i t th through h “TAG content“ t t“ Formulate multi-dimensional queries At Run/LB and Event level User Query caching Input Retrieve results (number of forms) Event counting g Display Results Stream Temporal Quality Trigger Physics Graphical tabular Retrieve events in ROOT file Register collection in AMI 30 Jan 2009 COUNT SELECT RETRIEVE events attributes events Elizabeth Gallas 8 https://voatlas18.cern.ch/tagservices/index.htm ELSSI Gateway Choose FDR2C li k link 30 Jan 2009 Elizabeth Gallas 9 ELSSI – 3 Frames: Banner, Session, Browser Banner ELSSI Browser Gateway Session 30 Jan 2009 Elizabeth Gallas 10 ELSSI – Useful buttons ELSSI Gateway 30 Jan 2009 Elizabeth Gallas 11 Main Frame: the Browser Frame 1. 2. 3. “Tabbed format” 1. Create Query • Temporal (run,time…) (run time ) • Streams • Data quality • Trigger • Physics attributes 2. Review Query 3. Perform Query • Count • Display results • Tabular • Plot • Retrieve event collection 30 Jan 2009 Elizabeth Gallas 12 Left Frame: the Session Frame Save/Share Session(s) S Save a query session i for later use Uses cookies Browser specific Share saved sessions with others Selection Summary y Displays selection criteria of your session 30 Jan 2009 Elizabeth Gallas 13 Summary y Metadata in a database is a powerful tool A great variety of dynamic selection criteria, criteria reporting and retrieval become possible Technical challenge (databases and ATLAS tools) Development is ongoing to improve Interfaces Functionality Usability Integrate I t t infrastructure i f t t ttools l For comments/suggestions: ATLAS Physics Metadata Hypernews 30 Jan 2009 Elizabeth Gallas 14 Backup Slides 30 Jan 2009 Elizabeth Gallas 15 TAG job with Pathena Panda Pathena jobOption ELSSI or POOL Collection Utilities Output Dataset events root events.root POOL Collection Utilities TAG DB events.root events.root events1.root DQ2 Catalogues DQ2 Client 30 Jan 2009 Elizabeth Gallas 16 Ganga g TNT 30 Jan 2009 Elizabeth Gallas 17 Working g on … ((1)) Incorporating Detector Status (DQ) Loading online Runs Even in the absence of TAG production An ELSSI “skim” skim button (Skimming Service) using GangaTNT: From your selected TAG events, generate an AOD dataset with only those events Output registered as new dataset Runs GANGA on the server to run the job on behalf off the h user User proxy is delegated to the server Monitoring g loop p checks jjob status and notifies user of job completion / failure Integrate GangaTNT into GangaGUI. 30 Jan 2009 Elizabeth Gallas 18 Working g on … ((2)) Using TAGs with PAthena: Steps create an event collection (events.root) with ELSSI setup Pathena and the Grid environment use a joboption like the one you'd use to run a local analysis based on TAGs submit the Pathena job use Panda monitoring to check for completion use DQ2 enduser clients (dq2 (dq2-get) get) to retrieve the output files Problem: AOD to AOD (copy) is problematic in some releases 30 Jan 2009 Elizabeth Gallas 19 Browser front page for FDR2 – ‘Temporal Tab’ 30 Jan 2009 Elizabeth Gallas 20 Browser – ‘Stream Tab’ 30 Jan 2009 Elizabeth Gallas 21 Browser – ‘ Trigger Tab’ 30 Jan 2009 Elizabeth Gallas 22 ‘Ph i Attributes ‘Physics A ib Tab’ T b’ 30 Jan 2009 Elizabeth Gallas 23 ‘Review Q Query y tab’ 30 Jan 2009 Elizabeth Gallas 24 ‘Perform Query’: y Count tab Count events meeting criteria … revise criteria… count again… revise criteria … 30 Jan 2009 Elizabeth Gallas 25 ‘Perform Query’: Display Results tab Get an idea of the attribute distributions Draw histograms g 30 Jan 2009 Elizabeth Gallas 26 Retrieve Events After refining query, generate a ROOT TAG file with selected events 30 Jan 2009 Elizabeth Gallas 27