[
https://issues.apache.org/jira/browse/OAK-1413?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13899303#comment-13899303
]
Michael Dürig commented on OAK-1413:
------------------------------------
At http://svn.apache.org/r1567690 I added
{{org.apache.jackrabbit.oak.jcr.LargeOperationIT#slowListener}}. This test
asserts that slow event listeners do not impact overall system performance too
much. The test registers 100 observation listeners, which process the incoming
events at half the speed new events are being created.
The results are quite encouraging: the test passes both for document and
segment node store. Even when half a million events are pending, system
performance does not degrade below the given bounds. Looking at the queueing
mechanism in
{{org.apache.jackrabbit.oak.spi.commit.BackgroundObserver#contentChanged}} I
found the commit queue to fill up quickly such that multiple commits coalesce
into a single change set. So overall this seems to work as intended.
> Add scalability tests for large operations
> ------------------------------------------
>
> Key: OAK-1413
> URL: https://issues.apache.org/jira/browse/OAK-1413
> Project: Jackrabbit Oak
> Issue Type: New Feature
> Components: jcr
> Reporter: Michael Dürig
> Assignee: Michael Dürig
> Fix For: 0.17
>
>
> We should add tests asserting large operation scale linearly in the order of
> the size of the operation. Candidates are:
> * large commit
> * move/copy of large tree
> * many siblings
> * large number of pending observation events
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
