we can deploy major upgrades on new hardware.
our failure plan for some non-critical systems, where perhaps up to an hour of occasional downtime is acceptable, is to monitor the box, and if it fails, spin up a new box and restore the system from backups
we can scale up in response to load events, rather than in advance of them
we can not worry about pre-launch capacity calculations
Virtual hardware doesn’t last as long as real hardware.
You need to be in more than one zone, and redundant across zones.
Multi-zone failures happen, so if you can afford it, go multi-region too.
(Full Story: AWS: the good, the bad and the ugly | awe.sm: the blog)