The booking portal of Nordlicht Reisen GmbH, a fictional travel agency, runs on a single virtual machine. If the machine fails or Azure restarts it for maintenance, the portal is gone. You are to make the machine more resilient and make sure that more machines run automatically when traffic is heavy. The exam puts it this way: deploy VMs to availability zones and availability sets and deploy and configure VM scale sets.
Prior knowledge. What regions, availability zones and redundancy are is explained in "AZ-900 compact", lesson 2.1; what virtual machines are, in lesson 2.3 of the same course and in lesson 3.3 of this course. This lesson shows how you align them for failures and load.
The building blocks of availability
You have to keep four terms apart:
| Term | Meaning according to Microsoft |
|---|---|
| Fault domain (FD) | group of hardware nodes that share the same power supply, network, cooling and maintenance schedule. VMs in different fault domains are probably not affected together by a planned or unplanned outage. |
| Update domain (UD) | group that is updated together during planned maintenance. Only one update domain is updated at a time. |
| Availability set | logical group of VMs that Azure spreads across multiple fault and update domains. |
| Availability zone | physically separate group of datacenters in a region with its own power supply, cooling and network. In regions with zones there are at least three. |
The effect: An availability set protects against hardware failures of a rack and against simultaneous restarts during maintenance. A zone additionally protects against the failure of a datacenter. Microsoft calls zones the highest reliability; availability sets remain vulnerable to shared failures, such as a network problem in the datacenter that hits several fault domains. In return, availability sets have lower latency between the VMs, because they are physically closer together.
| Availability set | Availability zones | |
|---|---|---|
| Protects against | failure of a fault domain, simultaneous maintenance | failure of a datacenter |
| Agreement (SLA) | 99.95% with two or more VMs in the set | 99.99% for instances spread across several zones (according to the scale set table) |
| Cost of the means | none for the set itself, only the VMs | check the price page for virtual machines |
| Combination | not with zones for the same VMs | not with an availability set for the same VMs |
| Occurrence | also in regions without zones | only in regions with zones |
The percentages come from Microsoft's documentation (as of November 11, 2026); the exact conditions and exceptions are in Microsoft's SLA texts.
Figures for sets: An availability set spreads across 2 to 3 fault domains (depending on the region) and up to 20 update domains. In Uniform scale sets there are 5 update domains.