Attending

• Tim Jenness
• Unknown User (xiuqin)
• Michelle Gower
• Russell Owen
• Dominique Boutigny
• Simon Krughoff
• Jim Bosch
• Brian Van Klaveren
• Unknown User (pschella)
• John Swinbank
• Kian-Tat Lim
• Unknown User (npease)
• Unknown User (kooper)

Agenda

• Overview of data access/databases/file-storage-systems in operations.
• Specify/split up and group/consolidate use cases.

Overview

• Michelle Gower presented an overview of the systems used in operations. These are primarily:
  
  • Operation production (archive, raw files, metadata in databases, etc.)
  • Operation L1
  • Release (qserv, read-only that external users query, provenance, metadata, tables, etc.).
• Michelle Gower, Jim Bosch, Brian Van Klaveren contributed additional data access systems:
  
  • Developer systems (subsets of `Operation production` used for developing and debugging, etc.)
  • Personal databases (accessible from Jupiterlab, live in qserv (probably the same instance as `Release` to enable cross-joins))
    • Brian Van Klaveren mentioned that creating and ingesting data into these user databases is not done by the user itself)
  • External local pre-cache
  • External remote on-the-fly
  • External large local (e.g. DESC@NERSC, HSC@Japan)
• Jim Bosch then suggested to go through the use cases and decide for each of them for which of the above systems it was relevant. A case from Simon Krughoff was selected as an example: "A science user sees a bad coadd and wants to find out what went wrong by inspecting the full chain of things that went into this coadd." This was further specified into:
  
  • 1.a Find inputs into coadd (warps, calexp, etc)
  • 1.b Get files into my sandbox (exists on disk in DB)
  • 1.c Get intermediate files that were used to generate this product (Brian Van Klaveren, Michelle Gower does this involve regenerating data on-the-fly?)
  • 1.d Inspect files by eye.

Additional discussion

• Brief discussion about need to read/write e.g. panda's datatable into user database (Jim Bosch yes, Brian Van Klaveren no).
• Discussion between Brian Van Klaveren, Tim Jenness, Jim Bosch and Michelle Gower about being able to reconstruct the qserv database from files on disk. Plan is to be able to do that (so qserv is not ultimate reference?).
• Kian-Tat Lim and Jim Bosch talked the about availability of qserv instances for development.
• Brian Van Klaveren, Unknown User (xiuqin) and Jim Bosch discussed the spectrum between regular users (expected to use IVOA interfaces to access data vs advanced users that will use the butler).
  

Conclusion and task for next meeting

• Go through each use case and:
  • split it up into specific steps,
  • then mark each step based on relevance to each data access system (using template matrix).