[
https://issues.apache.org/jira/browse/JCR-2699?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jukka Zitting updated JCR-2699:
-------------------------------
Attachment: ConcurrentReadWriteTest.png
ConcurrentReadTest.png
I've now implemented all of the above improvements (except ACL evaluation
changes, which are better handled in JCR-2573) and some other changes like
allowing concurrent reads on pooled bundle db persistence managers. Apart from
the ACL handling, I think I've now cleared all the major concurrency
bottlenecks above the persistence manager layer. Most notably a cache miss is
no longer enough to block all concurrent readers.
The attached plots show the current performance in comparison to previous 2.x
releases. The test case randomly accesses 1000 nodes out of a hierarchy of
10,000 nodes (100x100) while 20 other readers are also concurrently accessing
the nodes. In the read/write case a single random writer thread is also
included.
The lower bound (10th percentile) of all tests is at about 25ms, which
corresponds to the case where all accessed content is immediately available in
the item state or bundle cache. The median and the higher bound (90th
percentile) show the effect of one or more cache misses. The performance
improvements are about 20-25% compared to Jackrabbit 2.1.
Note that these tests were run with the default Jackrabbit configuration, i.e.
without access controls.
> Improve read/write concurrency
> ------------------------------
>
> Key: JCR-2699
> URL: https://issues.apache.org/jira/browse/JCR-2699
> Project: Jackrabbit Content Repository
> Issue Type: Improvement
> Components: jackrabbit-core
> Reporter: Jukka Zitting
> Assignee: Jukka Zitting
> Attachments: AccessControlProviderHolder.java,
> ConcurrentReadTest.png, ConcurrentReadTest.png, ConcurrentReadWriteTest.png,
> DynamicSecurityManager-JCR2699.patch
>
>
> I'd like to set up a few performance tests to help identify our worst
> bottlenecks for various kinds of concurrent read-only and read-write access
> patterns.
> Once identified, I'm hoping to fix at least some of those bottlenecks.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
