I paraphrase here, but in essence this is the message in this article and I would wholeheartedly agree.
Of course, the bit I want to highlight is to do with the discussion unfolding about bringing services back online quickly. I had to chuckle a bit when mention was made of bringing up a new instance of 2008 in milliseconds – I am not sure what this really referred to – certainly not a DR situation when you would typically have at least a 15 second delay before a fault would be detected in the first place – then you have to fire up a VM somewhere else, let W2008 go through its boot sequence, bring services online, maybe check a database before letting people back on – ummmm…….recovery time is building up here folks – users are offline!
Then there is a classic line that these solutions are not suitable for transaction processing.
On face value I would agree but then the author has not considered what is possible with fault tolerant technology – these systems have no single points of failure so no failover required.