ethbak commented on PR #16627: URL: https://github.com/apache/lucene/pull/16627#issuecomment-5608084377
Some benchmarking results: Cohere-1024, 500k normalized float32 vectors indexed as 5x100k segments, M=32, beamWidth=200, forceMerge(1), eight merge workers, Corretto 25, and 16 pinned CPUs on a Graviton desktop. ### 1% deletes | Arm | copy+repair | parallel insert | total | vs rebuild | |---|---:|---:|---:|---:| | reuse: before (#16618) | 11.2s | 122.8s | 134.0s | 0.86x | | reuse: this fix | 5.6s | 125.3s | 130.9s | 0.84x | | rebuild | 4.2s | 151.7s | 155.8s | 1.00x | ### 5% deletes | Arm | copy+repair | parallel insert | total | vs rebuild | |---|---:|---:|---:|---:| | reuse: before (#16618) | 26.5s | 120.1s | 146.5s | 1.00x | | reuse: this fix | 8.0s | 120.0s | 128.0s | 0.87x | | rebuild | 4.0s | 142.3s | 146.4s | 1.00x | ### 20% deletes | Arm | copy+repair | parallel insert | total | vs rebuild | |---|---:|---:|---:|---:| | reuse: before (#16618) | 112.2s | 99.2s | 211.5s | 1.73x | | reuse: this fix | 20.9s | 99.1s | 119.9s | 0.98x | | rebuild | 3.5s | 119.0s | 122.4s | 1.00x | ### 35% deletes | Arm | copy+repair | parallel insert | total | vs rebuild | |---|---:|---:|---:|---:| | reuse: before (#16618) | 97.5s | 79.3s | 176.8s | 1.74x | | reuse: this fix | 17.4s | 78.6s | 96.0s | 0.94x | | rebuild | 3.0s | 98.9s | 101.9s | 1.00x | -- 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]
