I guess the question needs to be restructured that instead of server uptime you would need SERVICE uptime. When we make apache upgrades, the apache service goes offline and restarts after the compile is completed. So while the SERVER may be powered on and working for 200+ days, that doesn't mean that the software did not kick out for a few seconds on an upgrade.
Many of the monitoring type sites that run the free reports scan on a 1 hour interval, or they are runing PING requests more than a service request such as HTTP, MySQL, FTP etc.
Of the 200 servers that I sampled earlier, our SERVICE uptime average is 99.923%. So there are a few servers where SMTP was as low as 99.8, but that doesn't necessarily mean that the website is offline.
99.9% uptime means approx 40 minutes of down time each month. I still think that's excessive for website needs! 99.99% (like some places state) means no more than 4 minutes of an outage.
What kills me are the hosts that claim 99.999% uptime, or even those saying 99.99% but they monitor their systems on 15 minute intervals, or they monitor on 5 minutes but don't take action until the 3rd alert (15 minutes later).
We run active monitoring pings every minute, on each service on each machine. 3 failed means 3 continious minutes of an outage which sends alarms, popup windows on staff's computers and pager alerts to staff.
That website that Steve posted says that Hands-on has had a 99.873% uptime since December 2005. The only problem with their reporting is that they are monitoring the main page of our site and not individual servers. Still, 99.8 over a 4 year period is good in my book
