[
https://issues.apache.org/jira/browse/OAK-1418?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13916280#comment-13916280
]
Jukka Zitting commented on OAK-1418:
------------------------------------
Revision 1573042 contains a minor performance improvement for this case, though
that change is probably not related to why the performance regressed in the
first place.
I profiled the benchmark to get the following breakdown of where most of the
time in a Node.getProperty() call is spent:
* 22% in ContentSessionImpl.checkLive()
* 18% in Tree.getProperty()
* 15% in RefreshStrategy.needsRefresh()
* 13% in getOakPathOrThrowNotFound(), including 10% in getSessionLocalMappings()
Based on that breakdown it looks like we still have a lot of room for further
optimization. The only thing that stands out as a possible cause of the
regression would be getSessionLocalMappings() in getOakPathOrThrowNotFound(),
as that's an area that I did touch earlier this month. I'll look into it.
> Read performance regression
> ---------------------------
>
> Key: OAK-1418
> URL: https://issues.apache.org/jira/browse/OAK-1418
> Project: Jackrabbit Oak
> Issue Type: Bug
> Components: jcr
> Environment: Oak 0.17-SNAPSHOT
> Reporter: Marcel Reutegger
> Fix For: 0.19
>
> Attachments: ReadPropertyTest.png
>
>
> A recent benchmark run of ReadPropertyTest shows a performance regression.
> The most recent run (206) was executed last night and the previous one (200)
> is from Feb 6th.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
