A Universal Framework for Data Validation

1
A Universal Framework for Data Validation

• Giovanni Organtini
• INFN-Sez. di Roma

2
Data Validation

• Data Validation (DV)
• Fast analysis of data coming from a DAQ system to
  evaluate data usability
• Each atomic analysis program is called task
• fast, light and single-objective
• Data Validation must be
• Able to evolve on a short timescale
• Flexible to manage multiple and different
  validation tasks
• Robust to allow many different users to cooperate
  without interference

3
DV Framework Concepts

• Sensors elements that detect events such as an
  EOR or a Warning signal issued by analysis tasks
• Events any Sensor outcome
• Mediator a process that collects Events from
  Sensors and dispatches them to Solutions
• Solutions objects that recognize Events and act
  according to their content

4
Framework architecture
5
Framework architecture
Sensors
6
Framework architecture
Notify Event to Mediator
7
The Sensors

• AlarmSensor parses HTML-formatted files and
  detects comments of the form
• lt!- SensorHTML
• Creates events according to comment content
• EORSensor parses a text file (runlist) and
  detects updates. A line is added to runlist at
  each EOR

8
Framework architecture
Dispatch Event to Solutions
9
Framework architecture
Solutions
10
The Solutions

• Program launch one or more programs in a given
  directory (repair known problems)
• E-mail send an e-mail with arbitrary text to a
  list of recipient
• Window show a popup window on a given terminal
  with arbitrary HTML-formatted text

11
A more formal view

• Implemented in Java
• portable
• distributed (RMI-server created to allow Sensors
  to run on remote processors)

12
How it works

• Sensors detect changes in files
• an EOR-Sensor can detect an EOR looking at the
  content of a given file containing the runlist
• An AlarmSensor can parse output files produced by
  analysis tasks to detect potential problems
• The Mediator collects Events and notify them to
  all the existing Solutions
• The Solutions choose to activate themselves
  according to Event content

13
EOR-Sensor

• The EOR-Sensor generate an Event as soon as a new
  line appear in the runlist
• runlist formatted as follows
• ltcommands/beam/gt --date20031111 ebeam50
  beame

14
EOR-Sensor

• The EOR-Sensor generate an Event as soon as a new
  line appear in the runlist
• runlist formatted as follows
• ltcommands/beam/gt --date20031111 ebeam50
  beame

directory containing analysis tasks for this
application
15
EOR-Sensor

• The EOR-Sensor generate an Event as soon as a new
  line appear in the runlist
• runlist formatted as follows
• ltcommands/beam/gt --date20031111 ebeam50
  beame

directory containing analysis tasks for this
application
parameters to be passed to analysis tasks
(arbitrary)
16
EOR-Sensor

• The EOR-Sensor generate an Event as soon as a new
  line appear in the runlist
• runlist formatted as follows
• ltcommands/beam/gt --date20031111 ebeam50
  beame
• Program Solution activated

directory containing analysis tasks for this
application
parameters to be passed to analysis tasks
(arbitrary)
17
Analysis tasks

• Analysis tasks can be written in ANY language,
  provided they can accept arguments on the command
  line
• Examples C, C, Perl, shell scripts, etc.
• They must provide the output in HTML-formatted
  files to be located in a given directory
• output can be either text and/or images
• the result of any single task could be browsed
  from everywhere using any browser
• a process will automatically generate the list of
  output files to be shown to the shifter

18
Alarm Sensors

• They look for HTML-comments in output files and
  generate events according to the comment content
• comments not shown by browsers
• must be compliant with a set of rules
• actions may be combined (i.e. can send e-mails
  and notify a message to the shifter)

19
Example

• lt! SensorHTML
• winFlagfalse
• emailFlagtrue
• commandFlagfalse
• ltemailBodygt
• senderEmailGiovanni.Organtini_at_roma1.infn.it
• senderNameGiovanni Organtini - ped analyzer
• receiverEmailtb.coord_at_cms.cern.ch,G.Organtini
  _at_cern.ch
• smtpServerlocalhost
• subjectProblem with pedestals in test beam
• ltmessagegt
• A problem has occurred during the analysis
  of the
• pedestal. May be you want to run to the
  detector to
• investigate
• lt/messagegt
• lt/emailBodygt
• ..--gt

20
Solution activation

• The e-mail Solution is activated
• sends an e-mail to Coordinator and Task
  responsible
• the e-mail body is arbitrary and may contain
  links and other HTML-tags
• Other informative message can be written on the
  output file containing the comment
• output file may contain
• text
• images
• links

21
Applications

• CMS Crystal Calorimeter Construction in Roma
• Crystals measured by automatic machine
• Results analyzed by validation tasks
• Problems notified to construction supervisor via
  e-mail
• Some known problems recovered via program
  Solution
• CMS Crystal Calorimeter Test-Beam at CERN
• Calorimeter modules exposed to e, p beams
• Data collected and analyzed by validation tasks
• Problems notified to experts via e-mail and to
  shifters via popup windows

22
Conclusion

• A Data Validation Framework has been developed
• simple to manage and use
• universal
• flexible
• robust
• configurable
• no requirements on analysis tasks
• no library to link or file to include
• Package available soon (2004) for download on
  http//www.roma1.infn.it/people/organtini