Gavin, where have you been?
I think Gavin is getting a little confused about service 'availability', 'reliability' and 'time to restore', probably because he wants to sell third-party managed cloud and hence be able to make warm, but woolly, statements...
Many small business's already operate with systems that are available 24/7 (in-house or cloud-based), largely because modern IT systems are surprisingly reliable! However, if and when anything goes wrong 'time to restore' is all they see ...
The challenge occurs where the business decides that it needs to avoid the outage downside, and sets an availability SLA of say 99.5%. Now we are talking about changing the IT architecture to support various degrees of redundancy and failure. It is here that costs can rapidly increase, with little to show for it. However, even this SLA doesn't include an explicit 'time to restore' requirement, which will have further ramifications on the solution and governance architectures.
The problem is that with many people now having been exposed to public cloud services, that are largely free and seemingly available and reliable - plus whilst the outages are frustrating at the time, because in the main they don't do anything really critical on them, work around the failures and soon forget about them (I suggest asking people about the recent MS cloud outage). Hence they will naturally question these costs and want to push stuff out to the cloud where they think they can obtain similar service levels but on a monthly subscription and hence avoid the capital outlay.
I disagree with one of Gavin's conclusions:
"what we need to be doing is blurring a solution’s non-functional availability and reliability requirements and be using statements like “the service will be available 24/7, but when there is a problem, whatever the problem, normal service will be resumed in less than 60 seconds.”
No as architects we should be seeking clarity, so that the business fully understands what it is getting for its money, it may decide that it is okay for it to take 3-hours to restore normal service when an outage occurs out-of-normal hours, rather than incur the costs of having a permanently manned facility/datacentre.
Aside: In my experience, the real problem I've had with iPads/BYOD is the whole MDM (Mobile Device Management) piece and the proliferation of client platforms. For example, enabling the finance director to look at KPI dashboards remotely isn't too much of a problem, with a laptop; but enabling them to usefully access the same applications on their iPad/smartphone etc. can be much more more problematic...