99.999% availability on service level agreements generally accomodate just one point of failure. So in a cluster if one machine goes down, it copes as there is another one to take it’s tasks on the fly and all the data is replicated elsewhere.
I wonder what High Availability(HA) setups feature two points of failure or how they’re marketed…
Be aware of 99.5% uptime guarantee perticulars. Most providers have a sliding scale of how much they compensate you. The compensation rarely adds up to anything significant.
If the total Downtime in the calendar month is more than fourteen and four-tenths (14.4) hours, the Service fee for that month shall be reduced by thirty-three percent (33%).
One third of your hosting costs for one month doesn’t quite cover the costs of a client’s business process that went down for an entire business day.
* Actually monitoring your server – outsource to Netcraft? :)
* DOS attacks etc.
* When dependent link on the net goes down that’s out of your control
* Insurance policies (urgh!)
* Power disruptions – most “UPS” solutions offers just a few seconds to shut down your server. Better setups feature generators, but they only function for as long as there is fuel. ;)
Keep dependencies to a minimum