Question 1, is can you have multiple un-clustered Integration servers installed on one physical box?
Question 2, is can you configure one of the integration servers as a reverse invoke server and the other as the internal server?
In answer to your Question 1, “can you have multiple un-clustered Integration servers installed on one physical box?”, you certainly can. Of course they’ll need to listen on different ports. This is actually a fairly reasonable technique if you’ve got a large box (e.g., 16 processors), because most JVMs can’t effectively use more than about 4 processors. So by running multiple instances of the IS, you can take better advantage of the hardware. It’s still (usually) cheaper to buy four 4-way boxes, than one 16-way, so this isn’t a particularly recommended answer, but I give it as an example of how your solution might be used.
In response to your Question 2 “can you configure one of the integration servers as a reverse invoke server and the other as the internal server?”, again the answer is yes, again given different port numbers. However, this time I’ll recommend against it… the main reason for having a reverse invoke server is to provide an additional layer of security to prevent direct attack on the machine running IS. But if they’re on the same machine, then it’s just adding a layer of complexity without really adding any security. If someone can compromise the RI server, then in they’ve also got the internal server, in the case proposed… while the general goal of RI is that even if you capture that server, you haven’t got the “crown jewels”.
Hope that helps!
–Jeremy