Just for fun this morning we reviewed Wikidot.com's uptime statistics over the last months. After fixing caching issues this summer, and before upgrading to the new super-fast database servers, we had about 99.7% uptime, and since then it's around 99.85%.
This comes down to about 1 hour downtime per month, which is caused by specific failures in different parts of the Wikidot cluster: usually due to extreme load peaks.
We're aiming for higher uptime (99.99% or under 5 minutes per month would be nice), in two ways. First, to improve the way we respond to problems, to detect them earlier and recover faster (and without service interruption). Second, to continue work on replication and caching so that your Wikidot.com sites have failsafe backups, both static HTML and live Wikidot.