On the 400 stuff we actually use JBoss, I use glassfish for our new projects which don't go out to the 400.
However on JBoss we use the IBM driver, and we create a new connection pool in the helper class we have, run the job close the connection and destroy the pool. (Horribly inefficient IMO but I didn't write the inital code)
I had a large select/update I ran and it used 84% of the CPU (on the 400) for a good half hour.
It shouldn't matter if you make a connection pool or not it should use the same server CPU time
[Message sent by forum member 'amattas' (amattas)]
http://forums.java.net/jive/thread.jspa?messageID=232765