1. Current system is at: http://glast-ground.slac.stanford.edu/SystemTests/
    • Histograms read from Root files via xrootd
    • Summary page displays status of series of "tests", with count of "failed" plots for each test
      • Success/Failure calculated at fill time by KS comparison with reference plot
        • Results stored in oracle "meta-data" table
        • Other statistical comparisons can be computed on fly ("statistics tab")
        • Can plot "time history" (actually version history) of meta-data from database
    • Plots displayed with overlayed reference plots (user can change default reference plots)
    • Tree for browsing individual plots or groups of plots.
      • Zoom in to any plot from page of plots by clicking on it
    • Admin functions allow
      • Plots to be "grouped"
      • Limited control of plotoing "style"
      • Hide old code versions
  2. Stategy
    • Should keep current system tests as "Software Release Tests"
    • Should keep good features of current system, but be flexible in evaluating different requirements of data quality monitoring
    • New system should bring together as much as possible of monitoring data in a single "web application"
      • Should link in to e-log,  housekeeping, activities, quicklook etc.
  3. Planned Features
    • Flexible method of choosing time period over which plots are displayed
      • Primary method of selecting plots will be run based, but should also support time period, orbit #, etc
      • Should be possible to integrate plots from multiple runs (may be required for some plots)
      • Probably similar mechanism to current trending application
    • Plots can come from histograms in root files, dynamically created from root tuples or from fits tables
      • Need to test performance for generating plots "on the fly" from tuples
    • Should be able to display tabular data as well as plots
    • Overlay of reference plots, or plots from different run/time ranges
    • Need way to display summary plots, plus drill down to see more details
    • Flexible method of testing plots for "success/failure"
      • Selection of algorithms for comparing with (expert selected) reference plots
      • Algorithms for testing individual plots (e.g. smoothness)
    • Interactive features
  4. Plan for January
    • Ability to display plots from root files on run by run basis
    • No reference plots initially
    • Need
      • histograms (and/or tuples) created by pipeline for each run
      • some database giving location of files (pipeline database may be sufficient)
  • No labels