[widebody911]

There is no magic number. I've personally managed just under 500 UX boxes, solo. It was a mix of AIX, Solaris, HP-UX and Linux, 80% of which was for application development and testing. My secret is that I automate anything that can be automated. The less human intervention, the better. I kept native OS images of all of the important boxes, and critical data was kept on the SAN. If something went south on a box, I'd lay the image back down on another piece of h/w and switch over - poor man's clustering.

The big wildcards are the number of users and types of apps on the boxes. Home-grown and custom apps tend to need more care and feeding than out-of-the box professional stuff. The more privs you give the users, the more problems they'll cause for you.

Another variable is the age of the boxes. If you're running 10 year old h/w trying to save $1, you'll spend $1.25 chasing down various h/w issues, but that won't show up on paper.

Microsoft stuff doesn't require as hefty a skillset to get up and running, but when something does go wrong, it's harder to fix not because of the actual complexity of the problem, but because the people administering the things usually don't understand them beyond "Right-click 'My Computer'..." For instance, the Verisign DNS boondoggle a couple years ago decimated our PC infrastructure, and the mouse jockeys were helpless because they didn't understand DNS. (Sure, they knew WINS...) The UX guys figured out what was going on and had a workaround established before the news officially broke.