[
http://jira.jboss.com/jira/browse/JBMESSAGING-406?page=comments#action_12... ]
Tim Fox commented on JBMESSAGING-406:
-------------------------------------
Currently messages are persisted via a RDBMs which is shared by all nodes in the cluster.
This provides a bottleneck. Also RDBMs are not well suited for persisting queue like
data.
Each node should have its own persistent store for performance and scalability.
For failover all persistent stores should available by all nodes on a shared file system
(e.g. GFS) - failover then becomes a simple matter of reopening the file based store on a
different node on failover.
Where a shared filesystem is not available we will have to consider replicating to a
buddy.
The best candidate for implementing local persistence currently is Berkely DB. However,
due to the Berkely DB licence we cannot distribute it with JBM, therefore it would be up
to the user to drop the Berkely DB jar in manually as part of the install.
We we would also need to provide a very simple JDBC PersistenceManager implementation to
allow to run the tests and play about with the default distribution - this can work with
HSQLDB.
Enabled local persistent storage per node
-----------------------------------------
Key: JBMESSAGING-406
URL:
http://jira.jboss.com/jira/browse/JBMESSAGING-406
Project: JBoss Messaging
Issue Type: Task
Reporter: Tim Fox
Assigned To: Tim Fox
Priority: Critical
Fix For: 2.0.0 Beta 1
Implement non shared file based persistent store.
Look at whether we can re-use stuff (ObjectStore) from JBoss Transactions to implement
this.
Also look at HOWL.
http://howl.objectweb.org/
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira