[
https://issues.apache.org/jira/browse/JCR-2950?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13401380#comment-13401380
]
angela commented on JCR-2950:
-----------------------------
ok... so i would suggest that we don't start changing yet another part....
i did run the ConcurrentReadAccessControlledTreeTest performance test that
should provide some results for
situations where the entry-cache limit is reached with JCR-3352 included and
with the 3 different cache update variants.
i get the impression that overall performance was similar in case of
synchronized and parallel access.
the throttled update looks to be slightly slower on average (but that
necessarily doesn't mean too much and i am not
convinced that the difference is really significant). what strikes me however
is, that with the 'T' strategy the performance
test seems to be stalling at some point.... but i have to check that one more
time... what also strikes me is that
the number of ms to read 10000 random items seems to vary much more in the 'T'
case (all completed cycles are
reported with system output even those from the warm-up phase).
in order to get a better feeling for that i added a shortcut in the beginning
of the "throttledUpdateCache" method that
prevent entering the concurrent map in case the target node is not access
controllable... but from the pure figures
this didn't seem to have an effect.
> CachingEntryCollector ineffective if number of accessed policies exceeds
> cache size
> -----------------------------------------------------------------------------------
>
> Key: JCR-2950
> URL: https://issues.apache.org/jira/browse/JCR-2950
> Project: Jackrabbit Content Repository
> Issue Type: Improvement
> Components: jackrabbit-core, security
> Affects Versions: 2.2.12, 2.4.2, 2.6
> Environment: Repository with ACEs > 1000
> Reporter: Honwai Wong
> Assignee: angela
> Attachments: CachingEntryCollector.ConcurrentCache-trunk.patch,
> JCR-2950-concurrent-cache-2.patch, JCR-2950-futures.patch,
> JCR-2950-futures_2.patch, JCR-2950-futures_3.patch, JCR-2950-futures_4.patch,
> JCR-2950-refactor+rootnode.patch, JCR-2950-refactor+rootnode_2.patch ,
> JCR-2950-refactor+rootnode_3.patch, JCR-2950-refactor+rootnode_4.patch,
> JCR-2950-refactor+rootnode_5.patch, JCR-2950-refactor+rootnode_6.patch,
> JCR-2950-refactor+rootnode_7.patch, JCR-2950-refactor.patch,
> JCR-2950-throttle.patch, JCR-2950-throttle2.patch,
> JCR-2950_performance_tests.patch.gz
>
>
> The CachingEntryCollector's cache (LRUMap, max size: 1000) seems to become
> ineffective in case there are more than 1000 ACEs present in the repository.
> Since access to the cache is synchronized, many threads are basically
> blocked, waiting to get access to the cache.
> Java callstack:
> at
> org/apache/jackrabbit/core/security/authorization/acl/CachingEntryCollector.getEntries(CachingEntryCollector.java:99(Compiled
> Code))
> at
> org/apache/jackrabbit/core/security/authorization/acl/EntryCollector.collectEntries(EntryCollector.java:134(Compiled
> Code))
> at
> org/apache/jackrabbit/core/security/authorization/acl/CompiledPermissionsImpl.canRead(CompiledPermissionsImpl.java:250(Compiled
> Code))
> at
> org/apache/jackrabbit/core/security/DefaultAccessManager.canRead(DefaultAccessManager.java:251(Compiled
> Code))
> at
> org/apache/jackrabbit/core/ItemManager.canRead(ItemManager.java:426(Compiled
> Code))
> at
> org/apache/jackrabbit/core/ItemManager.createItemData(ItemManager.java(Compiled
> Code))
> at
> org/apache/jackrabbit/core/ItemManager.getItemData(ItemManager.java:379(Compiled
> Code))
> at
> org/apache/jackrabbit/core/ItemManager.itemExists(ItemManager.java:292(Compiled
> Code))
> at
> org/apache/jackrabbit/core/ItemManager.itemExists(ItemManager.java:464(Compiled
> Code))
> at
> org/apache/jackrabbit/core/session/SessionItemOperation$1.perform(SessionItemOperation.java:49(Compiled
> Code))
> at
> org/apache/jackrabbit/core/session/SessionItemOperation$1.perform(SessionItemOperation.java:46(Compiled
> Code))
> at
> org/apache/jackrabbit/core/session/SessionItemOperation.perform(SessionItemOperation.java:187(Compiled
> Code))
> at
> org/apache/jackrabbit/core/session/SessionState.perform(SessionState.java:200(Compiled
> Code))
> at
> org/apache/jackrabbit/core/SessionImpl.perform(SessionImpl.java:355(Compiled
> Code))
> at
> org/apache/jackrabbit/core/SessionImpl.itemExists(SessionImpl.java:751(Compiled
> Code))
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
