[
https://issues.apache.org/jira/browse/QPIDJMS-553?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17504117#comment-17504117
]
ASF GitHub Bot commented on QPIDJMS-553:
----------------------------------------
franz1981 commented on pull request #45:
URL: https://github.com/apache/qpid-jms/pull/45#issuecomment-1063833632
FYI I've run this PR with the shiny new benchmarking tool in Artemis:
- using 100 producers/100 consumers with
`transport.sharedEventLoopThreads=-1` at the rate of 1000 msg/sec the RSS of
the process is 1.8 GB
- using 100 producers/100 consumers with
`transport.sharedEventLoopThreads=1` at the rate of 1000 msg/sec the RSS of the
process is 1.3 GB
Defaul linux java thread stack size is 1 MB and here we're saving 99 + 100
threads = 199 * 1 MB = 199 MB
The reason why the saving is even higher is probably due to Netty event loop
intrinsic cost (thread local(s), allocator(s) data structure(s), epoll OS
selectors, java Thread infra, task queues etc etc).
--
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]
> Shared Netty event loop group
> -----------------------------
>
> Key: QPIDJMS-553
> URL: https://issues.apache.org/jira/browse/QPIDJMS-553
> Project: Qpid JMS
> Issue Type: New Feature
> Reporter: Francesco Nigro
> Priority: Major
>
> One of the most interesting feature of Netty while using KQueue/NIO/Epoll in
> non-blocking mode is to be able to handle many connections with few threads;
> this is going to be critical and even more important with the upcoming
> IO_URING support, where the time spent on the Netty event loop to handle
> network syscalls will be further reduced, allowing syscall batching across
> different connections.
> Having the chance to handle many client connections with few Netty threads is
> already beneficial in constrained environments (containers with few cores) in
> order to reduce the native and heap memory usage.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
