I'm surprised that Jackrabbit would use a single connection to the
database.
Did you ask in the Jackrabbit list?
Also be aware that there are optimized Jackrabbit persistence
managers that exist now, they've been contributed recently to the
Jackrabbit project. They're called "Bundle" persistence managers:
Those pms combine the node and property states into a single bundle
and store them together. this improves performance and reduces
storage-memory overhead (no exact numbers available). The bundle
pms also have a "bundle-cache" that does a memory sensitive caching
of the bundles and a negative cache for non-existent bundles. small
binary properties are inlined into the bundle rather than stored in
the blobstore.
Documentation is light though. For an example, see
http://svn.apache.org/repos/asf/jackrabbit/trunk/jackrabbit-core/src/
main/java/org/apache/jackrabbit/core/persistence/bundle/
MySqlPersistenceManager.java
There's also some Javadoc in
http://svn.apache.org/repos/asf/jackrabbit/trunk/jackrabbit-core/src/
main/java/org/apache/jackrabbit/core/persistence/bundle/
AbstractBundlePersistenceManager.java
Florent
On 20 Apr 2007, at 15:20, Eric Kleinhans wrote:
To reduce the number of objects created on the file system by
Jackrabbit we plan to use a RDBMS backend for Jackbbit to store
metadata.
Looking at the source code of the Jackrabbit
SimpleDbPersistenceManager, it appears that Jackrabbit is using a
unique db connection to the RDBMS, which is ok for tests purpose
but can create a bottleneck in a production environment.
Has anybody already developped a patch for this pb (ex: using a
datasource/connection pool) ?
Eric.
--
Florent Guillaume, Director of R&D, Nuxeo
Open Source Enterprise Content Management (ECM)
http://www.nuxeo.com http://www.nuxeo.org +33 1 40 33 79 87
_______________________________________________
ECM mailing list
[email protected]
http://lists.nuxeo.com/mailman/listinfo/ecm
