Meeting began  @ 10am; ended 11:10am


Next meeting  @ 11am

The main topic for this meeting was estimating the effort and resourcing levels for implementation of the butler design developed as part of this Working Group.

Unknown User (pschella): Do we have enough information to estimate? Are we evolving the current new butler?

Kian-Tat Lim: Suggest new design but leveraging some of the existing code. Approximately 50% might be reusable.

Unknown User (pschella): How much effort was spent on original butler?

Kian-Tat Lim: May not be relevant. Need to identify the reusable 50%. New design is more complete and better understood than when first butler was developed.

Jim Bosch: Suggest we divide the work into components and estimate the components separately. We know the core functionality. Determining who implements plugins is tougher. Proposed design is in DMTN-056. Would like to list concrete entities needed for some use cases. For example, we know we need SQLite and Oracle support.

Michelle Gower: Worried that we are heading too deeply into design without having the requirements approved into official baseline.

Kian-Tat Lim: Design is general enough that functional requirements should be handled okay. Some concern with performance; especially with all the SQL joins. Also worried about physical decomposition of tables. Some of the concrete extensions listed on confluence are not distinct so may share a lot of interface code.

Unknown User (pschella): Plan is to talk to NCSA database admins to discuss performance issues for schema. We only have one design so hard to compare with other designs.

Jim Bosch: Can probably mock up year-N database with schema to test performance but that is a significant project in itself.

Michelle Gower: Do we need that by November? (No).

Kian-Tat Lim: Meeting with DB experts is good. Something that Kian-Tat Lim can also investigate.

Brian Van Klaveren: Do we know if this database will have SSD?

Michelle Gower: Not happy if the entire design mandates specific hardware.

Kian-Tat Lim: Need to show that we have some margin.

Unknown User (pschella): Have prototype implemented with SQLAlchemy and writing some tests. We will be using SQL database as backend. Will be hard to change later (no disagreement with SQL).

  • No labels