Project Meeting 2021.02.23

Technical Call

• Plan to get started on representative logsums soon
• Review issues
  • Jeff to fix #386 and look into #380
• Update from Doyle on tidying up agency examples
  • Now have PSRC, ARC, SEMCOG, and Marin in testable form:
    • Cropped so portable/runnable
    • Cleaned-up any issues with running with the latest code (and documented in a change log)
    • Have a test to test for successful model run
    • Travis now runs mini/cropped version
  • Built "cropper" tool
  • PSRC example exercises chunked accessibility code (but not for TVPB since PSRC is a two zone model)
  • Jeff to push code to develop2 and then share code with Stefan and Wu to test
• Update from me on the task 6 prototype multiple models test system
  • I'll keep working on it based on feedback and we'll discuss again next week
  • Part of the cost of AWS is automation to avoid excess costs
  • AWS great for common access
  • Need to use consistent language and terms
  • Could put on an AMPO server as well
  • Could amortize cost over annual budgets
  • Need to better define use of future warnings
  • AWS has an instance scheduler that PSRC is investing
  • We run mini/cropped version often and full version less often but at least with agency update
  • How much does the consortium want to be in the model correctness business?
  • There's multiple levels to this - does it run, is it perfectly correct, but also how's performance/scaling, and other in-between states
  • What's the use case to get new features into the core?
  • Let's discuss/add the ARC example to help explore the ideas
  • An additional goal is to make it easier for additional agencies to more easily adopt new methods
• Update from Newman on estimation work
  • See progress on issue #382
• Clint will share the ARC implementation next week