The case
The app is in São Paulo. The customer, in London.
Isaac's store started selling to Europe. The orders arrived. The customers' patience didn't.
This post builds on the concepts in Cloud Fundamentals. If elasticity and pay-as-you-go aren't clear yet, start there.
If you don't know the concept of latency yet and want to go deeper, read this post: Latency.
The problem
A single data center
The application works and serves Brazil well. But, as the store grows, two problems appear: high latency for those far away and a single point of failure.
Partial solution
A second zone
A zone is a data center (or a group of them) isolated within a region, with independent power, cooling and network. With two zones in the same region, if Zone A fails, Zone B keeps serving.
This solves the local failure, but not everything: both zones are still in São Paulo. If the whole region is affected, the application goes down, and the customer in London is still far away.
The solution
Multiple regions
A region is a separate geographic area, like São Paulo or London. With a second region in London, the solution is complete: low latency for Europe and resilience to zone failures and even to a whole region failure. If São Paulo goes down, London takes over.
Don't mix them up
Region × Zone
Region (region)
“What if the whole city stops?”
- Example
- A blackout hits São Paulo, and the London region takes over the traffic.
- Focus
- Disaster recovery, data residency and global latency.
Zone (zone)
“What if a data center stops?”
- Example
- Power fails in Zone A, and Zone B keeps serving.
- Focus
- High availability within the region.
| Region (region) | Zone (zone) | |
|---|---|---|
| What it is | Separate geographic area (São Paulo, London) | Isolated location within a region |
| Protects against | Losing a whole city | Losing one building or one power feed |
| Goal | Surviving a disaster, meeting local data rules, shortening the distance to users | Keeping the app up inside one region |
The name in each cloud
The concept is the same across all providers. Only the zone's name changes:
| Provider | Region | Zone |
|---|---|---|
| AWS | Region | Availability Zone (AZ) |
| Azure | Region | Availability Zone |
| Google Cloud | Region | Zone |
| Oracle Cloud | Region | Availability Domain (AD) |
In practice
How to choose the right region
The big clouds have regions spread around the world, and the same city often hosts regions from several providers.
Compliance
Some laws require citizens' data to stay within the country.
Latency (latency)
Choose the region closest to most of your users.
Service availability
Not every service exists in every region.
Price (pricing)
Costs vary between regions (local taxes, operating costs).
Quick test
Three questions, exam style
01
The application must stay online if an entire data center loses power. What is the minimum required?
02
The law requires that the data of Brazilian customers stay in Brazil. Which criterion decides the region?
03
An earthquake can take down all the zones of a region. What protects against that?
Regions and zones exist in every cloud certification: AWS, Azure, Google Cloud or Oracle Cloud. The name changes; the reasoning about latency and failure is the same.