[
https://issues.apache.org/jira/browse/JCR-2573?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
nihongye updated JCR-2573:
--------------------------
Attachment: acl.zip
i had focus the acl performance problem,and make a new acl implement,it work in
the way:
1.in many query,the result had same parent,to reduce collect times of parent
entries,i use LRUMap(100) to cache parentEntries
2.cache every ACE of ACLNode,a cache like ItemStateReferenceCache.see
CachedACLManager
3.avoid path operate,add a new method,public boolean grants(ItemId itemId, int
permissions,HierarchyManager hierMgr)
the test result show fast many times and small memory footprint
> Performance of AC Evaluation
> ----------------------------
>
> Key: JCR-2573
> URL: https://issues.apache.org/jira/browse/JCR-2573
> Project: Jackrabbit Content Repository
> Issue Type: Improvement
> Components: jackrabbit-core
> Affects Versions: 2.0.0
> Reporter: angela
> Assignee: angela
> Fix For: 2.2.0
>
> Attachments: acl.zip
>
>
> 1. Performance in access control evaluation
> =====================================================================
> - main focus on
> > read performance
> > resource-based access control in .a.j.c/s/authorization/acl/*
> - comparison admin vs. anonymous with full permissions
> - comparision between shortcut and ACL-evaluation.
> - comparison JR 1.4 vs JR 2.0 [actually i will compare Day's CRX as it
> already provided
> some custom AC stuff with JR 1.4]
> 2. Potential Problems
> =====================================================================
> I would expect the most significant problems to be found in
> a) ACLProvider#retrieveResultEntries: calculating effective ACEs
> for each session separately.
> b) AclPermission:
> Each instance registering an event listener in order to
> keep the result cache up to date
> c) AclPermission:
> Resolution form Path to Item or to nearest existing Item
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
