SincereXIA commented on pull request #616: URL: https://github.com/apache/ratis/pull/616#issuecomment-1061541975
Hi @szetszwo, Thanks for your opinion. I made a small change and tested your patch, and the result is here:  When unsafe flush + interval 3s is enabled, the average write performance of Ozone streaming is improved by 25% > Why async flush is slower then the single-thread flush? In the first version I implemented, I pulled out.flush() off the main thread and executed it every fixed time interval in a separate thread. So out.flush() was executed less than async flush. I guess that is the reason why async flush is a little slower. > Why async flush with current flushIntervalMin causing a bug? For this question, @kaijchen is right. Exactly, it's not a bug. In the last pr, I didn't consider that extending the refresh interval will cause the log submission time to be extended. The performance did not meet our expectations. In the current version, I use flushIntervalMin to control the flush interval of unsafe flush, which will not prolong the time that the log is finally committed. -- 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]
