[
https://issues.apache.org/jira/browse/JCR-2050?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Michael Dürig updated JCR-2050:
-------------------------------
Attachment: jcr.log.excerpt.txt
Log file excerpt from benchmark.
Most effect is visible for large trees. While fetching nodes is of the same
order of magnitude for both strategies, refreshing is remarkably faster for
LazyInvalidation.
> Optimize refresh operations
> ----------------------------
>
> Key: JCR-2050
> URL: https://issues.apache.org/jira/browse/JCR-2050
> Project: Jackrabbit Content Repository
> Issue Type: Improvement
> Components: jackrabbit-jcr-benchmark, jackrabbit-jcr2spi
> Reporter: Michael Dürig
> Assignee: Michael Dürig
> Priority: Minor
> Attachments: JCR-2050.patch, jcr.log.excerpt.txt
>
>
> With the current implementation (recursive) refresh operations cause a full
> traversal of the sub-tree rooted at the item causing the refresh. This is
> potentially expensive.
> Instead of invalidating each item in the respective sub-tree I propose to
> mark the root of the sub-tree as invalidated. Such a mark would include a
> time stamp. Also individual items would be time stamped with their resolution
> time. When an item is accessed, it would check if its resolution time stamp
> is older than the latest invalidation time stamp. If so, it checks whether
> the invalidation applies to it at all (by traversing up the path) and if so
> it would re-resolve itself. In any case its resolution time stamp will be
> updated.
> This approach would make invalidation much cheaper without putting much
> additional load to simple item access. Moreover most of the additional load
> (traversing up the path) only applies when an invalidation is pending.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
