Meeting began  @ 10am; ended 11:50

Attending

Next meeting:  @ 10am with Kian-Tat Lim, Unknown User (npease), John Swinbank and Fritz Mueller invited.

Design

Jim Bosch and Unknown User (pschella) talked through the design as seen in DMTN-056.

Questions for Kian-Tat Lim:

  • Is this design implementable?
  • Is the design lacking in specific areas?
  • Could the current butler support this design?

Some early comments from Kian-Tat Lim:

  • Consider Dimensions rather than DataUnits.
  • Worried about subsetting of the interfaces (batch production is inherently limited), configurability of the datastore, and reliance on SQL for the registry. Are you relying on the query optimizer? Is SQL sufficient?

Jim Bosch: Lots of scope for reusing current butler code for Datastore. Registry probably less so because of mapper rewrite.

Jim Bosch: Big problem is registry transfers.

There was some debate over whether staff could have limited write access to MyDB-like tables in DBB database to facilitate joins.

Kian-Tat Lim feels that 50% of the current butler codebase could be reused.

(Kian-Tat Lim had to leave at 11:05am)

Russell Owen asked why a Butler could only talk to a single collection when one of the use cases requires we can get different versions of the same dataset made by different software versions.

Jim Bosch replied that this is a limitation of the Supertask pre-flight rather than butler. May need to change the SuperTask design. Simon Krughoff should follow up on this to determine if his use cases require SuperTask.

Unknown User (pschella) requests that people check the design against the current requirements.

First step is for everyone to pick one of their use cases and explicitly walk through it with the current design.

  • Jim Bosch to create confluence page for concrete examples.

Michelle Gower offered to send round the DESDM operations schema.

  • No labels