[
https://issues.apache.org/jira/browse/HBASE-5943?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13482110#comment-13482110
]
Lars Hofhansl commented on HBASE-5943:
--------------------------------------
As Todd hinted, this isn't quite as clear cut as it looks, though.
Did some scan testing with various KV sizes and scanner caching settings, and
in some scenarios leaving Nagles on is actually faster.
> Set tcpnodelay as default
> -------------------------
>
> Key: HBASE-5943
> URL: https://issues.apache.org/jira/browse/HBASE-5943
> Project: HBase
> Issue Type: Improvement
> Reporter: stack
> Assignee: stack
> Fix For: 0.94.3, 0.96.0
>
> Attachments: nodelay.txt
>
>
> From IRC:
> {code}
> 14:46 < tlipcon> i turned off the FirstKeyOnlyFilter on the shell count
> command
> 14:46 < tlipcon> thgouhtput dropped like carzy... looked at strace, and saw
> 40ms delays
> 14:46 < tlipcon> turned on ipc.server.nodelay or whaever and it got fast again
> 14:46 < tlipcon> we should really ship that by default on
> 14:46 < tlipcon> there are probably people getting screwed by it
> {code}
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
