Yankee Group released a summary of their 2007 server reliability report. Among the findings:

  • UNIX systems continue to be near bullet-proof with over 99.99% uptime (IBM AIX lead the survey with 99.994% reliability .. about 30 minutes downtime per year).
  • Windows 2003 Server was less reliable, about 25% less reliable .. average downtime over 9 hours per year (99.9% uptime)
  • Linux on average was more reliable .. about 75% more reliable than last year averaging about 1 hour per year of downtime (99.99% uptime). This included both enterprise (paid) and community (free) variants of Linux. This was on-par with Solaris and HP UX.

Except for Windows 2003, the summary did not specify exactly what caused these outages. The summary specified many security updates requiring administrators to take down their Windows 2003 servers to apply the patches (its a shame you still have to take an entire server offline and reboot to apply a simple security patch!). It would be interesting to see if the outages for Linux were hardware related or software related. Most Linux installations that I am aware of are on consumer grade equipment which generally doesn’t have nice features to increase uptime numbers (ie hot-swap drives, RAID, redundant power supplies, etc) versus most mid-range and high-end Unix systems.