junegunn commented on PR #8001: URL: https://github.com/apache/hbase/pull/8001#issuecomment-4181566882
For those interested, here is a comparison with the previous work (#6557). See the `HBASE-29039` series without the `-alt` suffix. Both approaches effectively optimize read performance when delete markers are truly redundant: - <img width="2304" height="960" alt="image" src="https://github.com/user-attachments/assets/87b56c9f-777f-4d7e-8d9c-1ea73ff7f47d"; /> - <img width="2304" height="960" alt="image" src="https://github.com/user-attachments/assets/ec3783fa-4f26-40a2-897b-ba3367163ca2"; /> However, the previous work suffers from excessive seek overhead on false positives. This PR mitigates that overhead with qualifier comparison and an N=10 threshold. - <img width="2304" height="960" alt="image" src="https://github.com/user-attachments/assets/adc1ea77-9744-4320-954b-2aee9504928c"; /> - <img width="2304" height="960" alt="image" src="https://github.com/user-attachments/assets/649afab8-a1bb-452d-8742-9adafa00cfc5"; /> - <img width="2304" height="960" alt="image" src="https://github.com/user-attachments/assets/493e8304-e194-43a3-8ece-0e1c4944e168"; /> The full benchmark code can be found at https://gist.github.com/junegunn/bc0acf5269b8875330c0947dac7d0280. -- 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]
