franz1981 edited a comment on issue #2633: ARTEMIS-2317 Avoid long TTSP caused by Page::read using mmap read URL: https://github.com/apache/activemq-artemis/pull/2633#issuecomment-487370942 @qihongxu I'm getting similar results to yours :) This pr together with https://github.com/apache/activemq-artemis/pull/2646 should make numbers on par with the original version, but with an important difference: smaller time to safe point. TBH the "on par" statement need to be weighted: in the best case the mapped read is unbeatable 'cause it won't create garbage and will directly read from the OS page cache (on modern Linux boxes). But such performance comes with a side-effect: If you enable `-XX:+PrintGCApplicationStoppedTime` with the original version you will see longer pauses that could lead the broker to became unresponsive. It should happen with slow disk and/or OS page cache that won't contain the requested OS page (ie causing major page fault). The patched version should perform an additional copy into direct buffers, but it would perform read of the file using JNI, avoiding to be blocked by the OS on a major page fault. If you have any means to bench it in a similar scenario (with https://github.com/apache/activemq-artemis/pull/2646) it would be awesome! Many thanks for taking a look at it!
---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
