Rethinking High Availability

Have you noticed how the world has changed? Again.

A year ago I used to say that in public clouds three availability zones are enough. Each zone is in its own data center, data centers are located within ~100 km of each other, and the probability of losing two at the same time was considered very low.

There is no such assumption anymore.
Today, losing two or even all three availability zones in a single region is no longer extremely rare. It’s something that can actually happen.

What does it mean in practice?
If your business requires system 99.5% availability or higher, relying on a single region is not an option.

Technically it means:
- Using multiple regions even in public clouds (or even different cloud providers).
- Switching to cold or hot standby setup.
- Storing backups in a different region.
- Regularly testing DR scenarios.

Disasters do happen.
The question is whether your system is ready or not.

#engineering #systemdesign #reliability