https://bluejeans.com/426716450

Agenda

  • What's the story with terminology?
  • Table round of progress since last meeting. What progress has been made? What's blocking us?
    • Drill down. (Eric, Angelo, Tim)
    • Pipelines debugging. (Lauren, Simon, Trey)
    • Datasets & Testing. (Hsin-Fang, John)
  • Aims for next meeting.

Nomenclature

  • Unanimously adopted!

Drill down

  • Looking to use the interviews (see last week) to inform this narrative.
  • Need material on alert processing.
  • Lauren MacArthur compiled all the outputs from pipe_analysis; Angelo Fausti is going to look at these.
  • We could reimplement the scripts in pipe_analysis to use the verification framework.
  • Eric Bellm has been thinking about metric storage.
  • Next step is to unify all the thinking that the individuals in the group have produced.

Pipelines debugging

  • Simon Krughoff has thought about three separate scenarios in which you'd want to debug a pipeline.
  • Is lsstDebug really still on the menu?
    • Paul thinks so!
    • But getting it usable might be a lot of work.
    • And the mechanism is weird.
    • But we could replace it with something more idiomatic; “blue skies” (yuck) thinking.
  • Request to be able to plot things “survey wide”; this is exactly what Unknown User (tmorton)'s tools are doing.
    • Then subset data based on metric values.
    • (ie, find the visits with good metric values)
    • This argues for persisting metric values at the highest granularity possible.
    • What does that mean for data volume?
      • Eric Bellm has already been thinking about this.
      • Back-of-the-envelope estimate of CCD level metrics for 1 year of WFD survey is not overwhelming.
      • But we need to discuss how far we can push this at the source level.
      • Trading off latency on computation for storage.
  • Is there a database that we'll regularly be ingesting pipeline runs into, or is all the analysis to be performed on the data repository?
    • Probably some runs will go to a database, and some won't.
    • But it's in scope for the WG to express views on that.
    • It's not obvious whether storing metrics at the full granularity in a database is useful (vs other queryable formats)

Datasets & Testing


Next week's meeting

  • Each group come ready to present a summary of the picture they've arrived at to date to the group.