High Availability (HA) is provided by spreading backend services through multiple Availability Zones (AZ's). AZs are just like a datacentres. Having a single AZ provides a single point of failure, but you have a highly redundant infrastructure by replicating the services through multiple AZs. When one AZ goes down, traffic is diverted to the remaining AZs, giving you the impression that the world still has access to your system.
Call for advice now!