we've had very good luck with TCP clustering in production and in EC2 (which we use for dev and load testing) ... perf in EC2 seems to be more network bound, with sustained throughput topping out at 250 megabits for XL instances ... don't bother with the smaller instances as the bandwidth is less and the latency is higher. For 4 node clusters we were able to achieve roughly 90% of equivalent performance on real hardware (dual quad core at 2.0 GHz with 16 GB RAM) on EC2, though your mileage may vary. Overall, TCP is ok up to 8 node clusters n our experience (assuming gigabit) for read intensive apps. One other thing we did is to have dedicated bonded nics for the clustering traffic separate from client request traffic nics, and load balanced the VIPs across both the front and back planes. Hope this helps. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4173208#... Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...