I am fairly new to AWS and have designed a fairly flexible architecture which is fairly performant and fault tolerant over multiple regions/AZs
However I was thinking - what would the main disadvantages of sticking my entire stack (apps, DB, etc) onto one or two huge machines (r3.4x/8xlarge) or similar and then simply creating an AMI snapshot in S3 every x minutes and moving it cross-region to cover DR/HA
The performance of these machines (with provisioned IOPS/10 GbE) should be far in excess of my application scalability requirements for at least 3 years!