Modelling QA PR Tools

• Trying to form a functional model of Prompt Reco QA tools
• QA PR monitoring will consist of two parts:
  
  1. QaPrMon, which makes statistical tests of Dhp-format PR histograms, specified as comparison records;
  2. QaBrowser, which displays the inputs and results of comparisons.
  where 1) will run automatically after each new chunk of data is processed, and 2) runs in response to some user action.
• Components of QaPrMon mostly available through DhpXxx packages --- remaining questions concern interfacing and control.
• QaBrowser prototyped --- needs some additional functionality to suit it to user model.
• Foresee two user modes for QaBrowser:
  specific, in response to known problem with a given comparison;
  general, browsing a set of histograms produced over a given time period (and perhaps perform further comparisons)
• Two versions of the model are suggested:
  QaBrowser launched by ErrorBrowser, or
  QarRowser parallel to ErrorBrowser.
• Comments on these models are very welcome.

Some obvious questions about this model are:

• Which database are we talking about for storage of PR histograms?
• Could the requestor write directly to database?
• Is the "sleeping QaPrMon" a good idea, or should it be triggered by the writing of histogram objects to DB?
• In either case, how does QaPrMon know which histo objects to ask for and how to specify them to servers?
• Where is QaPrMon going to run? One of the nodes of the farm?
• Does QaPrMon need to write its results anywhere other than Error Logger? (e.g. for histories, fit parameters ...)
• Do live and reference histogram servers run all the time, or are they initiated by QaPrMon?
• How does the ErrorBrowser get enough information to launch QaBrowser with a specified comparison?

QaPrMon:

• Subsystems have asked for parametrized fitting as Dhp function; fits will be managed by QaPrMon.
• Would like statistical tests sensitive to individual parameters, so that error response different for shift in mean or change in width.
• Useful to keep fit parameters in DB for history, new references etc.
• .cri configuration file will become a database object.
• Need "offline" means for writing new references to DB.

QaBrowser:

• QaBrowser provides search facility to locate a given error in log.
• QaBrowser needs mechanism for user to specify histograms over time period, and means to merge them for display (end-of-shift check).
• May also want QaBrowser to initiate a comparison on merged histograms.
• Would be nice to provide link to web-based diagnostic information ... how can QaBrowser bring up a page in e.g. Netscape?
• Other simple functions include Print, display histogram statistics, bin contents etc.
• QaBrowser may obtain some of this functionality from JAS.