Revision progress tracker

[geoffxy]

New experiments

• [x] Run CH-BenCHmark on some fixed configuration and show BRAD's blueprint planner in action
  • [x] Add support for CH-BenCHmark (#486)
  • [x] Model updates
  • [x] Generate new training data based on the benchmark
  • [x] Update the GNN
    • [x] Implement and run data gathering
    • [x] Fit models
  • [x] Update the transaction models
    • [x] Implement and run data gathering
    • [x] Get to the bottom of the high transaction latency, decide on a path forward
    • [x] Fit models
  • [x] Update the hardware scaling models
    • [x] Implement and run data gathering
    • [x] Fit models
  • [x] Check in precomputed predictions
  • [x] Tune and select experiment scenario
  • [x] Plot results and add to revision
  • [x] Add TiDB baseline results to the plot
• [x] Scenario with changed SLOs
  • [x] Add ability to change the SLOs while BRAD is running (#489)
  • [x] Tune and select experiment scenario
  • [x] Plot results and add to revision
  • [x] Add TiDB baseline
  • [x] Update BRAD starting instance to improve cost differential obviousness
• [x] Re-run with table movement enabled
  • [x] Scale down
  • [x] Scale up (transactions)
  • Not needed to re-run; there will be no effect since the table placement does not change in the scenario
  • [x] Specialized functionality workload
• [x] Beam search sensitivity experiment

Writing changes

• [x] Add responses to reviewer comments
  • [x] Write polystore response (R1)
  • [x] Edit the baseline response (R1)
• [x] Writing adjustments
  • [x] Slightly alter introduction to better present the blueprint planning idea (requested by R4)
  • [x] Add short paragraph describing extensibility to additional engines (requested by R4)
• [x] Major writing adjustments
  • [x] Add more detail on the search algorithm
  • [x] Highlight physical plan novelty on the model, hoist this up more into the key ideas section
  • [x] Sprinkle in discussion points throughout the paper (Incorporate every comment into the main paper)

Nice to haves

• [x] Re-run scale down to have a larger transactional latency SLO buffer

[geoffxy]

This is complete!