The reason I ask is that we're investigating non-scalability of a standard Java benchmark (VolanoMark) on a Dell PowerEdge 2650 system (dual Xeon 2.4GHz). By "non-scalable", it is meant that there is no difference in results between smp kernel or not. btw, by "results" we mean the messages per second on the "loopback" test.
I've seen reports showing results that indicate that this benchmark should scale - well, at least some.
We just now tested on RH9, and, despite an obvious general increase in performance (Sun 1.4.2 jdk, too), there was no difference between using a single cpu or not (results were about 14,000 messages/sec either way).
TIA