Hey folks. Feel free to peek at the ActiveMQ message store implementations. this may be an area where our projects can cooperate. We have implemented JDBC based message store implementations that do batch operations. We have also implemented a message store filter that uses a journal like HOWL to do fast/concurrent syncs to a log and then do async checkpoints in batches to a journal. We have also started a BDB like file based implementation optimized for the messaging use cases that is still in development.
Some of the things we still need to do to our implementations is add better cursoring interfaces.. perhaps that some work we can do together? On 9/15/06, Robert Greig <[EMAIL PROTECTED]> wrote:
Just a quick note about persistence, which was discussed while I was away. As Rafi pointed out, we do have a generic interface that different persistence implementations can implement. Also note that the way it integrates with the configuration system means that the provider can add arbitrary configuration elements to config.xml and pick them up quite easily (e.g. JDBC url and driver would be required by a JDBC implementation). I think Colin hit the nail on the head when he mentioned HOWL. The reason I picked BDB originally was because it is essentially a high performance logger which I believe we need for performance reasons. I admit I have not benchmarked a straight JDBC implementation but I would expect it to be slower than BDB or (if it existed) a HOWL-based approach. As Rafi pointed out the implemtentation of a JDBC MessageStore would be relatively easily derivable from the BDB store implementation (in fact refactoring some of the BDB store into a common base class might be the best approach). Martin has also been working on getting the IBM JMS performance test harness to work with QPID, so if that is working we could hopefully be able to benchmark different approaches. RG
-- Regards, Hiram Blog: http://hiramchirino.com
