AWS tend to provide Multi zone and single zone availability and this is for each service they offer . So if you purchase multiple services ( say 5 ) but all are at Multi Region (99.9%) are you setting your application availability based upon standard ITIL practice of Multiplying the 5 targets together so you end up with 99.5% or because the likelihood if failure is small because you have built in multi zone capability you would set the target as 99.9% 

AWS appears to be more about zone availability that on premise hosting where you would base on servers and load balancing 

There are other factors you would add in like Network capability to get to the cloud hosting but am more interested in the cloud approach people are taking