At one of my previous jobs, the engineers on my team had an informal competition as to who could rack up the longest uptime on their workstation (they all had Sun Solaris or Linux, of course). When the company moved to a new office, one crafty engineer managed to beat all the others by putting his Sun into the seldom-used hibernation mode to preserve his uptime when everyone else was forced to reboot.

I posit that uptime is actually a bad thing. All software has bugs, and a regular maintenance schedule to apply patches, at the very least once a month, should be part of the plan and designed into the architecture. By that token, an uptime greater than 31 days is a “code smell” for infrastructure.