qihongxu commented 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-487358556 @franz1981 Thanks for contribution to page performance issues! I have deployed a test as soon as I saw this thread. Our configuration: 4C/8G/500G SSD single broker as usual 200 threads of consumers to squeeze out all potentials Most options in broker.xml remained default except for changing journal-type to MAPPED Then we tested its performance in two versions (one with this patch and the other not) Producers and consumers ran separately with each message sizing 4k. TPS is as below: | Producer | Consumer -- | -- | -- Without patch | 13k | 15k With patch | 13k | 14.8k It seemed that consumer tps in two versions were very close. We also tried raising page size from 10M to 100M, raising message size form 4k to 12k, etc. But none of these made a significant difference in tps between two versions. Flame graph for version with patch is as below:  @franz1981 Would you please give some hints on the differences between this and @clebertsuconic 's result? Is there any configuration that we missed in tests?
---------------------------------------------------------------- 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: [email protected] With regards, Apache Git Services
