Project Meeting 2021.03.23

Technical call

• Discuss SANDAG 3 zone testing and support
  • Need to spend funds by mid April so this is a short term priority, which is fine and good for everyone
  • Will focus on the two major slow items - accessibility calculator and location choice models, specifically trip destination choice
  • RSG has setup and run SANDAG's setup now. Accessibility calculator ran in under 2 hours. Our machine has 32 cores and 512 GB ram
  • We will review the setup and likely help optimize the num processors and chunking settings
  • Best to target a num processors that is about 80% of the physical processors instead of the logical processors (at least on Windows)
  • Jeff now setting up a cropped version for testing/debugging and will add example_sandag to the repo and then start profiling
  • SANDAG's current model runs in 6 hrs
  • Chunking can be configured for each step so can use different sizes for different submodels
• Discuss potential input checking and error handling improvements
  • I reviewed the potential list of input checking and error handling items and asked folks to review and comment
  • Will change the pipeline table persistence item to a "add an optional global switch to turn off restarting" for application mode
  • I built the list based on feedback from the SEMCOG development team as well
  • Let's try to at least pick a few of these items within the next couple of weeks since Jeff will plan to work on these after we help Wu with the performance issues
• Discuss updated improved test system plan next time
• Discuss also writing samples for location choice models for estimation data bundles next time