Github user franz1981 commented on the issue:
https://github.com/apache/activemq-artemis/pull/1361
@clebertsuconic There is no need to use a ThreadLocal or a pooled buffer
there because:
1) the method (write/read) is synchronized and it is safe to reuse the same
ByteBuf wrapper between different threads (note: is only a wrapper on the
ByteBuffer provided by the file factory)
2) the file factory already provides a (thread local) pooled direct
ByteBuffer (on NIO)
In the old version the most of the garbage was already avoided due to the
fileFactory ByteBuffer thread local pooling, but I've measured after multiple
Page writes lots of minor GCs due to many short living ActiveMQBuffer
instances. This change aims reduce that instances to be only 1 per Page: it is
not optimal but is a big improvement.
With 100 bytes messages and a Page of 10 MB it avoids about 104856
instances of garbage.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
