AZ-104
Deep Dive
Azure offers multiple options to guarantee that VMs stay available during hardware failures, planned maintenance, or datacenter disasters. The AZ-104 tests Availability Sets, Availability Zones, and their impact on SLA.
Contents
| Configuration | SLA | Max downtime/month | Conditions |
|---|---|---|---|
| Single VM with Premium SSD | 99.9% | ~43.8 min | OS disk and data disk on Premium SSD or Ultra |
| Single VM with Standard SSD | 99.5% | ~3.6 hours | OS disk on Standard SSD |
| Availability Set (2+ VMs) | 99.95% | ~21.9 min | At least 2 VMs in the Availability Set |
| Availability Zones (2+ VMs in different zones) | 99.99% | ~4.38 min | At least 2 VMs in 2 different zones of the same region |
| VMSS (Flexible, multi-zone) | 99.99% | ~4.38 min | Instances distributed across multiple Availability Zones |
For the exam — the three key SLAs
Memorize: 99.9% (single VM + Premium SSD) → 99.95% (Availability Set, 2+ VMs) → 99.99% (Availability Zones, 2+ VMs). Each tier raises the SLA by protecting against a different layer of failure.
Fault Domain (FD)
A Fault Domain is a group of hardware that shares a power source and network switch. If there's a hardware failure (power, networking), it only affects the VMs in that FD.
• A typical datacenter has 2-3 FDs
• VMs in the same Availability Set are automatically distributed across different FDs
• Maximum of 3 FDs per Availability Set
• A failure in FD1 → FD2 and FD3 keep running
Update Domain (UD)
An Update Domain is a logical group of VMs that Azure updates/reboots during planned maintenance. Azure never reboots two UDs simultaneously.
• Azure updates one UD at a time, waiting 30 min between UDs
• Up to 20 UDs per Availability Set (default is 5)
• While one UD is under maintenance, VMs in other UDs keep serving traffic
• If you have 2 VMs in 2 different UDs: at least 1 is always active during maintenance
An Availability Set groups VMs and distributes them across different Fault Domains and Update Domains within the same datacenter. It guarantees that not all the VMs fail at the same time.
How to configure it
Limitations of Availability Sets
Managed Disk alignment
With Managed Disks in an Availability Set, the disks are also distributed across different storage fault domains. Without this, multiple VMs in different FDs could have disks on the same storage rack and fail together.
Availability Zones are physically separate datacenters within an Azure region (each zone has its own independent power, cooling, and networking). Distributing VMs across zones protects against a full datacenter failure.
Characteristics
• Most Azure regions have 3 AZs (some have 2, some still have 0)
• Zones are ≥1 km apart from each other (latency <2ms between zones)
• When creating a VM: explicitly select Zone 1, 2, or 3 (or "no preference")
• Zone-redundant services: the Azure service is automatically distributed across zones (e.g. ZRS Storage, SQL Hyperscale)
• Zonal services: you choose which zone to deploy to (VMs, disks, IPs)
• Intra-region latency between zones: <2ms (round trip)
Availability Zones vs Availability Sets
| Aspect | Av. Set | Av. Zones |
|---|---|---|
| Protects against | Rack/PDU failure | Full datacenter failure |
| SLA | 99.95% | 99.99% |
| Scope | 1 datacenter | 3 distinct datacenters |
| Requires | 2+ VMs | 2+ VMs in different zones |
Mutually exclusive
A VM cannot be in an Availability Set AND in an Availability Zone at the same time. Choose one or the other when creating the VM.
What they are
A PPG groups VMs on the same rack or physically very close together, minimizing network latency between them. Unlike Availability Sets (which spread VMs apart), PPGs bring them together.
• Typical latency within a PPG: <1ms
• Without a PPG in the same region: latency can be 2-5ms
• Compatible with Availability Sets and Availability Zones
When to use a PPG
Trade-off
Lower geographic availability. Concentrating VMs on the same rack increases the risk that a single hardware failure will affect all of them.
What they're for
Capacity Reservations guarantee that Azure has compute capacity reserved for you in a specific region and zone. Without a reservation, during periods of high demand (regional disasters, massive events) Azure might not have VMs of the size you need available.
With reservation
• Azure guarantees capacity for N VMs of size X
• You are billed even if you are not using the VMs (you pay for the reservation)
• Ideal for DR — you know capacity will be there when you fail over
Without reservation
• You pay only for active VMs
• No capacity guarantee at critical moments
• VM creation can fail if the datacenter is full
| Option | Protects against | SLA | Extra cost | Replaces |
|---|---|---|---|---|
| Premium SSD (single VM) | Disk failure | 99.9% | Cost of Premium disk | — |
| Availability Set | Rack failure / planned maintenance | 99.95% | Free | N/A (legacy) |
| Availability Zones | Full datacenter failure | 99.99% | Free + cross-zone traffic | Better than AS |
| VMSS Flexible multi-zone | Datacenter failure + automatic scaling | 99.99% | Free for HA, cost of added VMs | Availability Sets |
| Proximity Placement Group | Latency between VMs (brings together, doesn't separate) | Variable | Free | — |
| Capacity Reservation | Lack of capacity at critical moments | Capacity guarantee | You pay even if you don't use the VMs | — |
Did you get this topic?
Test what you just learned
A company has 3 VMs that make up the web tier of their application. They need to guarantee the app stays available during Azure's planned maintenance and during rack hardware failures. Which configuration should they use?
Put what you learned to the test with original AZ-104 exam questions.
Start AZ-104 mock examUnlimited mock exams with the Pro plan
Log in to track your progress.