[
https://issues.apache.org/jira/browse/OAK-1757?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13977441#comment-13977441
]
Michael Dürig commented on OAK-1757:
------------------------------------
The actual CPU overhead seems to depend on the environment quite a bit. I've
seen everything between 4% and 10% and now you report 1%. This is confirmed by
the fact that I didn't see much of this using a JVM profiler. Only when falling
back to a system profiler on the JVM itself I started seeing the overhead of
frequent thread scheduling.
While I think 1% is certainly acceptable, 10% is IMO not.
On a more general note the changes from
http://svn.apache.org/viewvc?view=revision&revision=1575050 are not that
helpful for gathering correct sessions statistics. The 1ms resolution of the
timer is too low as many of the operations take sub milliseconds to execute.
> Oak eats too many CPU cycles when idle
> --------------------------------------
>
> Key: OAK-1757
> URL: https://issues.apache.org/jira/browse/OAK-1757
> Project: Jackrabbit Oak
> Issue Type: Bug
> Components: jcr
> Reporter: Michael Dürig
>
> An idle Oak eats roughly 5% of my CPU cycles.
> This is caused by the {{Clock}} implementations introduced at
> http://svn.apache.org/viewvc?view=revision&revision=1575050. Scheduling the
> fast clock to update its time every millisecond is quite expensive.
> Substituting for the simple clock reduces idle CPU usage to below 1%.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
