[
https://issues.apache.org/jira/browse/DIRMINA-1129?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17188733#comment-17188733
]
Jonathan Valliere edited comment on DIRMINA-1129 at 9/1/20, 6:49 PM:
---------------------------------------------------------------------
[~elecharny] it is important to set the SO_SNDBUF in order to prevent
unnecessary tcp buffer compaction because the OS defaults can allow up to huge
buffers and the process wastes a bunch of cpu cycles by allowing the
user-application to write a lot to the buffers in one shot but the
expand/compact is really expensive. Here is a thread where I discuss one of
the problems of NOT setting SO_SNDBUF:
https://unix.stackexchange.com/questions/304372/high-softirq-when-using-rate-control-networking
was (Author: johnnyv):
[~elecharny] it is important to set the SO_SNDBUF in order to prevent
unnecessary tcp buffer compaction because the OS defaults can allow up to huge
buffers and the process wastes a bunch of cpu cycles by allowing the
user-application to write a lot to the buffers in one shot but the
expand/compact is really expensive.
> Unable to decrease socket receive buffer size below 65535 for NIO socket
> connector.
> -----------------------------------------------------------------------------------
>
> Key: DIRMINA-1129
> URL: https://issues.apache.org/jira/browse/DIRMINA-1129
> Project: MINA
> Issue Type: Bug
> Components: Transport
> Affects Versions: 2.0.22, 2.1.4
> Reporter: Marcin L
> Assignee: Jonathan Valliere
> Priority: Minor
> Fix For: 2.1.5
>
>
> NioSocketConnector sets only SO_RCVBUF when configured receive buffer size is
> above 65535. Under some circumstances it might be desirable to set it to
> lower value.
> Can the check be replaced with the similar one to NioSocketAcceptor?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
