TheR1sing3un commented on PR #6625: URL: https://github.com/apache/rocketmq/pull/6625#issuecomment-1565590140
Sorry for busy these time~ Now I finish a optimized batch read method in `TieredFileSegmentInputStream` and its child class: `TiredCommitLogInputStream`. I also do a JMH benchmark between original batch read method in `InputStream` and new optimized method in `TieredCommitLogInputStream`. In a common scenario: - `msg size` : 4KB - `tieredStoreGroupCommitSize` : 32MB - `batch read size` : 8KB(In S3 client, it will read 8192 Bytes each time calling `read(byte[], int, int)`) > ops(read 32MB)  > average time(read 32MB)  **I think this optimization can improve performance by an order of magnitude** > for more different combinations of `msg size` and `batch read size` <img width="731" alt="image" src="https://github.com/apache/rocketmq/assets/87409330/78d24413-665b-4ef9-ab21-1d02e735a69c";> @zhouxinyu @ShadowySpirits -- 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]
