Isn't the standard way to implement clustering to implement a clustered JMS and then make everything else rely on that?
Kind of yes.
FWIW we've an Apache 2.0 licensed clustered JMS here...
http://activemq.codehaus.org/
which supports auto-discovering clusters of clients & message brokers and a variety of clustering options...
http://activemq.codehaus.org/Clustering?refresh=1
Also for a clustering implementation. Let's say I have information on Server A and I want to ensure it is clustered so I send it to server B. Is a standard implementation to fire it over a socket and wait for receipt confirmation and then move on or are people generally more sophisticated then that?
Its kind of a long story, but the short answer is, we're more sophisticated than that. Here's how distributed queues / topics work in ActiveMQ...
http://activemq.codehaus.org/How+do+distributed+queues+work?refresh=1 http://activemq.codehaus.org/Networks+of+Brokers?refresh=1
Dain Sundstrom wrote:
That module has basically been abandoned. I expect us to use ActiveCluser for the clustering.
Agreed.
We're already starting to use ActiveCluster as a way to discover & connect ActiveMQ message brokers to form a cluster of clients and brokers with HA. Also WADI - the clustered session state project - uses ActiveCluster as its discovery, communication and cluster protocol abstraction...
http://wadi.codehaus.org/
James ------- http://radio.weblogs.com/0112098/
