2020 11 17.devel

• Travis
  • changes to paid model
    • we have to limit workers
    • we are currently not affected, can observe limits
    • also, paying is ok really
  • same anaconda (but not condaforge)
  • github is expected to remain free for academia
  • we still have no real solution for data storage
    • github-lsf / cloud / self-hosted / ECE-hosted?
• TODO AGENDA RCT Nov. 17
  • data archival
• RP-#2248
  • TODO IP: check for reproducer on SAGA layer
• RA: experiences
  • RA is a programming module, not a collection of scripts
  • the scripts are useful, but are not designed to be the main entry point, but rather examples
  • maybe move scripts to a separate repository?
  • just using scripts does not give insight without careful understanding of plot semantics
  • maybe simpler version which gives a user level information
    • allocation + tasks (CPU + GPU) (no individual contributions)
    • task concurrency
    • higher level API?
    • notebooks as templates?
    • maybe look again into dashboards? Can be post-mortem
    • TODO ALL: what information do we actually want?
      • both for publications and quick session inspection
    • we need to reconsider how we do UID mappings and event caching
      • move from UID to list of UIDs
  • PRRTE profile pre-processing difficult to inline... :-/
  • TODO AM: RE meta data, object hierarchy
• RE: release

TODOs

• resources
  • list all valid resources in docs?
    • TODO IP: check script
    • TODO AM: check resource configs (tacc / xsede)
• TODO AM: move radical-base to sandbox
• TODO AGENDA: switch from VE to CONDA?
  • conda freeze?
• TODO AGENDA: async workflow in MIDAS
  • example: double loop around bag of tasks