[cle-release-team] SiteCacheImpl binding directly to Ehcache

David Haines dlhaines at umich.edu
Wed Apr 9 09:05:07 PDT 2014


If I remember correctly the separate maps for caching were added long ago
for performance issues.  The ehcache event handling was done because the
maps would not be cleaned up when ehcache itself evicted an entry.  That
led to lots of orphaned entries and a very significant memory leak.

Getting rid of all of that is great if the replacement is performant for
both time and space.

- Dave


On Wed, Apr 9, 2014 at 10:23 AM, Aaron Zeckoski <azeckoski at unicon.net>wrote:

> I am working on removing the direct bindings to Ehcache from the
> various caches in Sakai core and moving the kernel (and the rest of
> the core) towards JSR-107 compliance. As part of this effort I noticed
> that some caches use Ehcache directly (like
> org.sakaiproject.site.impl.SiteCacheImpl). For this case it looks like
> it was mainly to get access to Ehcache notifications via
> CacheEventListener. This appears to be because some of the cached data
> is replicated into a series of maps (m_tools, m_pages, m_groups) which
> then have to be manually cleared when the data is flushed.
>
> That code originated in https://jira.sakaiproject.org/browse/KNL-652
>
> This strategy of partly caching some data in the local object in maps
> is incompatible with modern caching systems and distributed caching.
> In order to move Sakai towards better support for scaling this
> strategy has to change.
>
> I created a ticket for this:
> https://jira.sakaiproject.org/browse/KNL-1162
>
> Before I remove the maps and Ehcache bindings (which will likely
> change the behavior/performance profile of this cache) I wanted to see
> if anyone had thoughts or comments about why this was done in the
> first place (goals, specific perf issues, etc.)?
> -AZ
>
> --
> Aaron Zeckoski - Software Architect - http://tinyurl.com/azprofile
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://collab.sakaiproject.org/pipermail/cle-release-team/attachments/20140409/1110715c/attachment.html 


More information about the cle-release-team mailing list