Meta%20Data - PowerPoint PPT Presentation

About This Presentation
Title:

Meta%20Data

Description:

Today new DataSets can be created only by a specialized application. Vincenzo Innocente, CERN/EP ... Clone (at a given deep level) a DataSet from one Owner to another ... – PowerPoint PPT presentation

Number of Views:58
Avg rating:3.0/5.0
Slides: 11
Provided by: ygap
Category:
Tags: 20data | meta | setup

less

Transcript and Presenter's Notes

Title: Meta%20Data


1
Meta Data
An Introduction to non-Event Data in CARF
  • Vincenzo Innocente
  • CERN/EP/CMC

2
Top Level Event Structure (ORCA4)
Run
Crossing
Trigger
Pile-up
SimEvent
3
Re-Reconstruction Clones
Run
Run
Id-1
Local Replica
Crossing
Trigger
Pile-up
4
Dataset Collection
MetaData User Tag
Run Collection
Rec Event
5
Top Level Meta Data Structure
System Collection
Owner
RunList
Specific to DS type
DataSet
Run
SetUp
EVDFilePool
Event Collection
Persistent Algorithms
EVDFile
Configuration
Specific to DS type
Container
Specific to DS type
Location of event data
6
Top Level Factories
  • Collections and Configurations depends on the
    type of event the dataset is composed of.
  • DataBase files (location and name) depend, on a
    lesser extent, on some given strategy
  • A CARF specialized application instantiates the
    correct factories using a specialized
    DBPopulator.
  • Today new DataSets can be created only by a
    specialized application

7
Easy, not fully implemented
  • List all OWNER in federation
  • grep oodumpcatalog
  • List all files in Owner or Dataset
  • grep oodumpcatalog
  • Delete a complete Owner
  • Eradicate a complete Dataset
  • Delete ALL data corresponding to a given run
  • Stop a batch-job before its natural end (without
    killing it)
  • Restart a batch-job from the point it was stopped
  • Clone (By ref or by value) a collection to a
    different owner

8
Need more work
  • Share EVD files among multiple DataSets
  • Create Owners DataSets outside ORCA
    applications
  • Pre-allocate EVD files (on different servers)
  • Write to different servers
  • Determine if a Running job is actually crashed
  • Determine that a running job is in a loop and
    stop it without killing it
  • Determine (efficiently, a.s.a.p. and for sure!)
    that no job will ever write into a given file

9
Need Schema Changes
  • Get History at Run Level
  • Get Principal Event-Id at MetaData Level
  • Store non-CARF metadata (Pythia, G3) in DB

10
Need development (and RD)
  • Clone (at a given deep level) a DataSet from one
    Owner to another
  • Share configuration among different Owner
Write a Comment
User Comments (0)
About PowerShow.com