The AA has decided to bring its datacentre operation back in house to save money. This isn't very surprising if you stop to think about the basic economics, but it does seem quite brave of them to admit that the outsource decision was flawed. It has always seemed unlikely to me that you could save money by outsourcing infrastructure in this way, unless your in house setup was totally incompetent (in which case fixing it would seem to be the better option). Unlike in the early days of facilities management, there is no option to save money by consolidation workload onto fewer mainframes, so the cost model for the outsourcer is likely to be the same as yours - add in a profit margin and account management overheads and you are paying more. Even worse is the degree to which your ability to support your business is compromised. I have worked on accounts with some large infrastructure outsourcers over the past year and the striking thing is how long it now takes to make any changes - and how much it costs.
The other interesting thing about this news item was the decision to accept a lower level of availability - 99.5% instead of the 99.9% or more that is usually asked for. Like Google's practice of running their datacentres rather hotter than is considered normal, it suggests that there are a few bits of conventional wisdom that could do with revisiting. All it takes is a clear look at the numbers
