Project Meeting 2024.07.25

Agenda

Admin

Phase 9a wrap-up status

• Usage docs basically done, pending re-review
• Need to add performance reporting summary to docs as future reference
• End support for dynamic chunking? Support a transition to explicit chunking instead?

Phase 9b kick-off

Meeting Notes

Admin

• September 30 is drop dead date for AMPO contracts
• Tentatively planning for SANDAG presentation of their model at AMPO conference

Phase 9a Wrap-up

• PR for documentation with sections on multiprocessing, multithread, using sharrow, using skims (effective data formatting) is effectively done. David and Sijia reviewed, and Jeff incorporated their feedback. They can re-review, but looks close to done and ready to be approved and merged.
• Jeff is going through various recent issues and checking to see if past relevant issues related to performance have been addressed. Jeff thinks he's close to done on that.
• Jeff will be consolidating performance issues/results documented in different issues into the final documentation.
• In going through recent issues, a couple of things where people are still having trouble with chunking. Optimized dynamic chunking system doesn’t work very well, and we can use explicit chunking effectively enough to get to get good performance results. Jeff proposes that explicit chunking is the supported approach. We can help you make explicit chunking work for you, instead of dynamic chunking. (Help those partners today that are using dynamic chunking and transition them.)
  • How would we document that we’re not supporting dynamic chunking?
    • Jeff can mark in the documentation that it’s a deprecated function. We wouldn’t strip the code; but by calling it “deprecated,” it would deter future users from using it. Once we are comfortable that all dynamic chunking is removed from current work streams of partners, we can strip from the code.
  • What would be the level of effort to switch existing dynamic chunking users to explicit chunking?
  • We are not sure how many implementations out there use dynamic chunking, but Jeff doesn’t think it’ll take much effort to transition.
  • Jeff will update documentation so that it’s very clear what you need to change to move to the explicit chunking.

Phase 9b kick-off

• RSG/Driftless Labs has two estimation mode tasks
  • Contracting is still in process
  • In two weeks, they can present their efforts and discuss in more detail
• WSP has explicit representation of working from home
  • Original scope includes adding components to have the software ready to be able to identify telecommuters, that agencies can be use for their estimation or calibration efforts specific to their implementation. The software code would be backwards compatible; the new variable could be not used.
  • The outstanding question is whether or to what extent calibration would be a part of this effort. The task order language hasn’t been settled.
  • WSP to describe proposed approach, along with limitations and recommendations next week. We'll also discuss the implications of requiring re-estimation and/or calibration. We'll discuss as a group next Thursday.