- scaling up vs. efficiency
  - single query: great if you can scale up
  - possibly expensive if concurrency
  - hardware/software/financial limits
  - distributed systems
- Use-Cases: HTAP/OLAP, not OLTP
- Goals:
  - explain where we're expecting to make progress
  - explain if improvements could help you
  - get feedback
- three fields:
  - same query plan faster
  - find a better, already executable, plan
  - new executor possibilities
- Past:
  - same query plan faster:
    - effective_io_concurrency
    - changes in 10:
      - expression evaluation
      - hashtable improvements
  - better plans:
    - false correlation, zip code & city ex
  - new executor possibilities
    - index only scans
- Future:
  - some query plan faster:
    - JIT expressions/deforming
    - executor refactoring
  - same query plan faster:
    - better stats / selectivity estimation
    - CTE inlining
  - new executor possibilities:
    - lazy hash join