Ref discussion above. Is it the industry or the application that determines availability needs? (Actually it's the user that determines importance!). Here is an interesting example. Beijing constructed an underground roadway to ease surface congestion, connect parking lots, and improve access to 2008 Olympic event venues. Abundant data collection devices and video cameras report back to a control center to enable rapid response to conditions and mishaps within the 5.5 km loop. This traffic management and control software is defined by the environment to be mission-critical, even if only for the duration of the event. Similar considerations can affect a virtualized infrastructure when, at certain times during the day, week or month, a particular application can become mission-critical. For that period of time the application is migrated to a server resource pool that is fault-tolerant.

I think virtualization could be driving more mainstream consideration of continuous availability and fault tolerance. Putting day-to-day apps running in a virtual environment alongside a number of other virtual environments on one machine tends to make that machine a mission critical piece of the business. So, the calculus of how reliable is reliable enough begins to change, IMHO.

I wonder if fault-tolerance is really only relevant in obvious industries (the banks, the airports and so on). I'd be surprised if some standard enterprise would go fault-tolerant just to support day-to-day apps.

Compared to alternative solutions like clusters, FT carries a pretty hefty premium. Even basic servers are incredibly reliable, so I have to wonder if fault tolerance is really worth that price difference.