M. Oldenburg GridPP Metadata Workshop July 47 2006, Oxford University 1

1
ALICE metadataOverview

• Markus Oldenburg
• GridPP Metadata Workshop
• July 47 2006, Oxford University

2
Run and File Level Metadata

• AliEn (Alice Environment)
• distributed computer environment for Alice
• provides
• Database interface (MySQL)
• File Catalogue
• Metadata Catalogue
• other services
• File Catalogue
• acts as and looks like a File System
• doesnt own the files, just associates logical
  file names with physical locations
• Metadata Catalogue
• file and directory structure reflects structure
  of underlying database
• additional tables can be attached to each
  directory ? metadata
• directory structure is chosen in a way to group
  similar files together
• reduction of metadata for a given directory
• enhancement of search performance

3
Event Level Metadata

• raw data is processed right after data taking
• some physical quantities will be extracted right
  away
• multiplicity
• vertex position
• each file containing physics events gets an
  additional file containing this event level
  metadata attached
• ? tag file
• root file
• stored in the same directory as the physics data
  file
• content can be extended later (or each user can
  even create his/her own tag files)
• actual analysis runs only over those events
  selected by certain cut criteria
• create a file list from the file catalogue
  (run/file level metadata)
• read tag file to select only interesting events
• loop over data

4
What is working so far?

• Everything!
• File Catalogue (with defined directory structure)
  exists and works)
• run and file level Metadata Catalogue (data
  fields) is defined and exists
• event level metadata is defined
• all stages were tested and work properly
• But
• no large scale tests yet
• many tables/catalogues not filled yet (at least
  not automatically)
• not enough simulation data to effectively stress
  test the system
• Currently
• large test production running
• output files are added automatically to the file
  catalogue
• system performance to be seen ?