1. Fault domains

• common set of hardware that has a SPoF like a rack

2. Update Domains

• group of nodes that are upgraded together

2. Availability sets (99.95%)

Protection against rack level failures.

1. Logical grouping of nodes so that they are deployed over different racks.
2. So that if 1 goes down, other is available
3. Don’t mix functionalities. Example 1 for DCs. 1 for sql for example.

3. [[202404081830 Azure Availability Zones|Availability Zones]] (99.99%)

Racks live in a DC set (separate power, cooling, network) This provides protection again DC level failures. Minimum of 3 zones in every region. Even if there are more, in your subscription you will see 3.

4. Regions and Pairs

Set of DC sets becomes a region. 2ms latency roundtrip window between DC sets/ Availability Zones. Paired regions - main thing is azure does not update both regions at the same time

---

references:

https://learn.microsoft.com/en-us/training/modules/configure-virtual-machine-availability/5-review-availability-zones