Real world scalability examples
I'm trying to get an idea about how well Glassfish (v4 in this instance) scales in real world use - both vertically and horizontally.
The use case I'm looking at is Glassfish hosting soap and/or restful services, serving a sustained rate of about 100 concurrent transactions/second (target is about 5-10M transactions a day at peak rates). In my simulated tests so far, if I raise the thread pool to around 512 threads (similar with concurrent connections, EJB pool, etc), I appear to be able to get this on a high end desktop. I'd like to know how well this translates to something like a four-socket, 32-core server, etc, and then to ten of these machines. Is it better to have fewer more powerful servers, or to go down the virtualization route and have lots of smaller systems?
I've had a quick look around, but haven't really found much. Does anyone have anything they can share?