apoorvmittal10 commented on PR #23911: URL: https://github.com/apache/kafka/pull/23911#issuecomment-6340935601
Ran some performance tests through claude to find if it doesn't cause any regression. Results: | Scenario | Metric | Trunk | This change | |---|---|---|---| | **Throughput** (3M records preloaded, consumed at full speed) | records/s | 249,667 / 266,785 | 268,192 / 255,450 | | | MB/s | 47.6 / 50.9 | 51.2 / 48.7 | | | broker CPU, consume phase (s) | 14.8 / 14.3 | 14.0 / 14.4 | | **minBytes** (producer capped at 5,000 rec/s, `fetch.min.bytes=1MB`, `fetch.max.wait.ms=500`, 300k records) | records/s | 4,973 / 4,996 | 4,980 / 5,017 | | | broker CPU, total (s) | 28.4 / 25.7 | 28.3 / 22.6 | | **Segment roll** (`segment.bytes=1MB`, ~50 segments per partition, 1.5M records produced and consumed concurrently) | records/s | 128,733 / 126,657 | 126,957 / 137,665 | | | MB/s | 24.6 / 24.2 | 24.2 / 26.3 | | | broker CPU, total (s) | 15.6 / 16.2 | 15.9 / 15.1 | | | Throughput | minBytes | Segment roll | |---|---|---|---| | Partitions | 6 | 6 | 6 | | Consumers | 8 | 8 | 8 | | Records | 3,000,000 | 300,000 | 1,500,000 | | Data | 572 MB | 57 MB | 286 MB | | Producer | Preloaded first at full speed, consumer started afterwards | Capped at 5,000 records/s total, running concurrently with the consumers | Full speed, running concurrently with the consumers | | Consumer overrides | none | `fetch.min.bytes=1048576`, `fetch.max.wait.ms=500` | none | | Topic config | defaults (1 GB segments) | defaults | `segment.bytes=1048576` (~50 segments per partition) | | Run length | ~11–12 s consuming | ~60 s | ~11–12 s | -- 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]
