Amazon’s cloud crashed a few days ago. We were not affected, but only because of a lucky bit of scheduling. As it turns out, we completed migrating to a more sedate host the day before the outage hit as a result of our pivot.
AWS is a great platform for just about any startup in general and for our sync project in particular. The offered scalability and the price for companies in our situation (free) is kind of hard to beat. But that price tag runs on a clock. We don’t need AWS while we are doing customer validation, so we switched.
Don’t get me wrong; the outage won’t keep us from going back to AWS when we launch our ATS. It will, however, cause us to dual host on AWS and another completely distinct, shared-nothing provider. Perhaps Rackspace, as a quick example. While the business case for a true 5 9′s (99.999% uptime) operational model is not quite there (trust me, for each “9″ you add, it’s usually at least an order of magnitude in operational costs — if not two or three), I do believe our customers will be expecting 99.999% uptime during general business hours.