I was thinking about that - at Mastercard we had some primitive system level thing that was half in the Java world and half in the daemon world, that diviied up the E10000 processors for the bloated threads of the massive Java app we were running - I wasn't privileged enough to know all the details..I do remember that clustering was in what I considered to be an amazingly primitive state compared to how it worked on DEC equipment. I did the failover clustering and it was a mess - the actual process level clustering was probably even more of a mess. (circa 2001)