What it is
An Availability Zone is a data centre, or a cluster of them, inside a cloud region. AWS and Azure both use the idea. The region is the city. The zone is the building.
The point is physical separation. Different power. Different cooling. Often different flood plains. If one building has a bad afternoon, the others in that city are supposed to keep working.
A region with one zone in use is still one building. The label on the invoice does not change that.
Why it matters in the meeting
When someone says “we are in us-east-1,” Bart will ask which zone, and how many. City is not the same as surviving a building problem.
JJ hears “highly available” and thinks the cloud will handle it. The cloud handles it if you paid for more than one building and actually put something in the second one.
Real world
The famous outages are often one zone having a bad day while neighbouring zones are fine. Workloads that lived in a single zone waited with everyone else. Workloads that spanned zones recovered, or at least stayed up.
Azure availability zones work the same way with different names on the map. The meeting question does not change: how many buildings, and what happens when one goes dark.
A zone is a building. A region is a city. If everything important lives in one building, you have a very expensive single point of failure with good branding.
What to ask
- How many Availability Zones does production actually run in, not just have access to?
- If zone A disappeared at 2pm, what would still answer?
- Are we treating a single-zone database as highly available because the slide said multi-region?
- Do AWS and Azure production workloads follow the same zone rule, or only one of them?
You just knew a little more Jack than you did five minutes ago.
All concepts