Someone (I think it was @Ray) asked about SLAs back in the timesharing days. The only one I remember (well) is the old Control Data commercial offering, called CYBERnet. This was classic timesharing and the SLA was simple – they guaranteed a particular availability level, and there were severe penalties (i.e. refunds/cash payments to the user) if that level was not met.
Many an employee of CYBERnet was therefore tasked with finding ‘whose fault it was’ for outages – the root cause being tagged as either ‘customer’ or ‘provider’. This was no easy task, of course, but it did result in a relationship of trust between the user and CYBERnet as time and effort was actually spent chasing root cause. On the inside at CYBERnet, there was an extreme importance placed on change control, as you can imagine, since unplanned downtime was so expensive.
But there were also planned outage (i.e. defined unavailable) periods during which the user could not use the service. This was built-in to all contracts. A typical outage period would be 8 hours over a weekend, once a quarter (just an example). Back in the day, it was called preventive maintenance.
The world of cloud storage providers is much different these days – there is no SLA of any sort, the provider (e.g. S3) declares themselves blameless in all cases, takes no responsibility and provides no penalties or guaranteed availability. They can also change fees anytime for any reason, and you have no resource except to cancel, and they kick in automatically after 15 days like it or not.
No soup for you.
Rob