The concept of AZs is really easy – in the case of a zone failure, your workload should still be intact. This means that AZs can secure you from a single data center failure (network outage, power outage), but still cannot protect you from an entire region disaster. If we assume that a region contains three zones and you deploy seven machines, the placement will look like this:
- Zone 1: VM1, VM4, VM7
- Zone 2: VM2, VM5
- Zone 3: VM3, VM6
If Zone 1 fails, you still have four machines able to work on your workloads. This does not mean that an application will not be intact (it will require some load balancing of the current services and possibly that you rerun some of the processes), but it can survive a possible outage.
On the other hand, AZs are meant to protect you only against a local failure – they do not guarantee durability when a disaster occurs that affects all the zones at once. To secure against that kind of failure, you will...
