andygrove opened a new pull request, #2109:
URL: https://github.com/apache/datafusion-ballista/pull/2109

   # Which issue does this PR close?
   
   This is a documentation refresh and does not close a specific issue. It 
records
   [#2025](https://github.com/apache/datafusion-ballista/issues/2025) (Q18 
build-side
   OOM at SF1000) as a known result at the new sizing.
   
   # Rationale for this change
   
   The benchmarking page's results were from an older reference cluster (2 
executors ×
   8 cores) and reported two Ballista configurations (AQE on and AQE off). 
Going forward
   these benchmarks use a single official configuration — 2 executors × 16 
cores, one per
   node, `target_partitions=64`, AQE on — so the page needs to reflect that 
shape and
   carry a fresh result set from it. Reporting only AQE on keeps the page 
focused on the
   planner Ballista is actively developed against.
   
   # What changes are included in this PR?
   
   - **Reference cluster** updated to 16 cores per executor (16 task slots, 3 
GB pool
     each) and `target_partitions=64`.
   - **AQE off removed**: dropped the second Ballista column and the "two 
planners are
     always measured" discussion; the page now measures AQE on only and says so.
   - **Result table refreshed** for Spark, Comet, and Ballista at SF1000, 1 
iteration,
     with pinned versions (Ballista `main` @ `49d1fec8`, Spark 3.5.3 vanilla, 
Comet
     `main` @ `b0165552`). All three engines plan `SortMergeJoin`, so the Comet 
run pins
     `spark.comet.exec.replaceSortMergeJoin=false` to match Ballista's join 
strategy.
   - **Q18 recorded as `OOM`** on Ballista: at 16 task slots sharing the 48 GB 
pool the
     non-spillable hash-join build side exceeds the container limit (#2025). 
The total row
     therefore sums the 21 non-Q18 queries so it stays comparable across 
engines.
   - Run commands and the "Recording a result" guidance updated to match 
(partitions,
     cores, AQE-on wording).
   
   # Are there any user-facing changes?
   
   Documentation only; no code or API changes.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to