Rangsh commented on PR #12173: URL: https://github.com/apache/seatunnel/pull/12173#issuecomment-5579204433
@SEZ9 interim evidence update (also posted on #12063): Fork `Benchmarks Diagnostics` stayed queued 36+ minutes without starting, so I cancelled that backlog. I cannot dispatch on `apache/seatunnel` (403). Collected a controlled **local** before/after on the same machine (JDK 17, `-f 3 -wi 3 -i 5`, `-prof gc`) for both methods and both counts: | Benchmark | Count | Before Score / Err% / CV% | After Score / Err% / CV% | | --- | ---: | --- | --- | | `runningJobGrowth` | 0 | 186.4 / 22.3% / 20.9% | 157.2 / 33.9% / 31.7% | | `runningJobGrowth` | 1000 | 197.8 / 14.3% / 13.3% | 124.8 / 5.7% / 5.3% | | `completedJobHistoryGrowth` | 0 | 1146.1 / 9.0% / 8.5% | 1041.0 / 11.4% / 10.7% | | `completedJobHistoryGrowth` | 1000 | 1623.1 / 54.4% / 50.9% | 839.8 / 7.9% / 7.4% | GC time/alloc dropped substantially at `count=1000`, especially completed. Local absolute scores are not comparable to the Actions baseline. After includes fixture + production, so attribution is not isolated yet. Happy to split to fixture-only if that is preferred while waiting for apache-side same-runner diagnostics. -- 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]
