[
https://issues.apache.org/jira/browse/JCR-2762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12920249#action_12920249
]
Jukka Zitting commented on JCR-2762:
------------------------------------
There's a few other optimizations I've been working on, most notably improved
handling of date and long values, but patches 1-4 already achieve most of the
potential size reductions. With these patches the size of typical bundles is
down about 30-40% from before. The main benefits of the smaller bundle size are
reduced disk usage and therefore faster I/O and more optimal use of disk caches.
Note that all persistence managers necessarily have some storage overhead over
the plain bundle size, so the total improvements will not be nearly as big as
suggested by the bundle size reduction. For example, with these patches the
size of the default workspace of a fresh installation of our proprietary CQ
product drops from 154MB to 141MB (-9%). The time it took to fill the workspace
(including writes to data store, etc.) went down from 85s to 72s (-16%), though
this was just a single try so the performance numbers are fairly uncertain.
I didn't yet commit these changes since changing the serialization format
always carries some risk (though we now have a fairly good serialization test
cases) and prevents switching back to an older version after an upgrade. The
question therefore is, should we do this now for Jackrabbit 2.2 or postpone
this for 2.3? I'm inclined to do this already in 2.2.
> Optimize bundle serialization
> -----------------------------
>
> Key: JCR-2762
> URL: https://issues.apache.org/jira/browse/JCR-2762
> Project: Jackrabbit Content Repository
> Issue Type: Improvement
> Components: jackrabbit-core
> Reporter: Jukka Zitting
> Priority: Minor
> Attachments: 0001-JCR-2762-Optimize-bundle-serialization.patch,
> 0002-JCR-2762-Optimize-bundle-serialization.patch,
> 0003-JCR-2762-Optimize-bundle-serialization.patch,
> 0004-JCR-2762-Optimize-bundle-serialization.patch
>
>
> There are a number of ways we could use to make bundle serialization more
> optimized. Thomas has already done some work on this in the Jackrabbit 3
> sandbox, and I'd like to apply some of the optimizations also to the trunk.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
