10
btw, a good buddy of mine is an IBM lead tech, he's responsible (among others) for keeping the IBM sites/networks up and running properly, and when they have millions of hits *PER MINUTE*, they've learned to keep every piece of the servers separate....
database on one, webserver on another, etc etc... and i believe they use distributed load (clusters), as well as having fallback servers in case of something going wrong..
btw, they use html, they've found it's the quickest and most reliable method for serving large numbers of online users at one time (less strain on the servers)