Lesson 09 — Virtual Machines: Configuration, Sizing & Availability

Compute Is the Heart of the AZ-104 Exam

At 20–25% exam weight, the compute domain is the largest single domain in AZ-104. Virtual machines are where most Azure administrators spend the majority of their day: provisioning, resizing, configuring disks, managing availability, deploying extensions, and troubleshooting. The exam tests both theoretical knowledge (which SLA does an availability zone give you?) and operational judgment (when should you use a spot VM? what disk type for a database?)

This lesson covers every major VM concept you need: VM families, the disk ecosystem, the four availability layers, proximity placement groups, spot VMs, extensions, images, and hybrid licensing. Master this and you have locked in over a fifth of your exam score before the next domain.

VM Families and the Naming Convention

Azure has dozens of VM series, each optimised for a specific workload class. The AZ-104 exam does not require memorising every series, but you must know the major families and be able to decode the VM size naming convention.

Decoding the VM Size Name

Understanding the naming convention is a practical skill and sometimes tested directly. The format is:

Additional letter modifiers you may encounter: a = AMD processor, d = local temp disk included, i = isolated (dedicated physical host), l = low memory variant, m = memory optimised variant within the family.

VM Creation Components

Deploying a VM is not a single-resource operation. A VM deployment touches multiple Azure resources, each with its own lifecycle and configuration options:

• OS Disk: Managed disk containing the operating system. Sourced from a marketplace image, a custom image, or an uploaded VHD. Type must be selected at creation (Standard HDD/SSD, Premium SSD v1/v2, Ultra).
• Data Disks: Additional managed disks attached for application data. A VM can have multiple data disks (limit depends on VM size). Each has independently configurable disk type and caching.
• NIC (Network Interface Card): Every VM has at least one NIC with a private IP in a VNet subnet. A public IP can be attached. NSGs can be associated at NIC or subnet level.
• Admin Credentials: Windows VMs use username/password. Linux VMs should always use SSH key pairs — password authentication on Linux should be disabled.
• Availability Configuration: Availability set, availability zone, or none. Must be configured at VM creation time — you cannot move a deployed VM into an availability set or change its zone later.
• Extensions: Agents and scripts installed post-deployment for monitoring, encryption, configuration management.

OS Image Types

Azure VMs are deployed from images. Understanding image types is important for VM fleet management at scale.

Azure Compute Gallery

The Azure Compute Gallery (formerly Shared Image Gallery) provides enterprise-grade image management:

• Image definitions: Metadata container specifying OS type (Windows/Linux), publisher, offer, SKU, and generation (Gen1/Gen2).
• Image versions: Actual image artifacts with version numbers (e.g. 1.0.0, 1.1.0). Each version can be replicated to multiple Azure regions for performance.
• RBAC sharing: Share images with specific subscriptions, tenants, or make them publicly available.
• Regional replication: Azure replicates image versions to chosen regions automatically, reducing deployment time for VM scale sets and large fleets.

Managed Disks

Always use managed disks. Unmanaged disks (page blobs stored in a customer-managed storage account) are a legacy concept — you have to manage the storage accounts yourself, availability set placement is limited, and Azure Backup integration is poor. Managed disks are standalone Azure resources with their own resource ID, independent lifecycle, and first-class Azure feature support.

Disk Types Compared

Disk Caching

Each managed disk has a configurable caching mode, which determines how the host server's RAM and CPU are used to buffer I/O before it reaches the actual disk:

Availability: Four Levels of Protection

Azure's availability model is layered. The level of protection you need — and the SLA you can claim — depends on which option you use. These concepts are tested heavily on AZ-104 because the exam loves questions that distinguish between the options and their specific SLA percentages.

Fault Domains and Update Domains

These are the building blocks of Availability Sets:

• Fault Domain (FD): A physical rack in a datacenter that shares a power source and network switch. VMs in different fault domains will not be simultaneously affected by a rack power failure or network switch failure. Availability sets use 2–3 fault domains.
• Update Domain (UD): A logical grouping of VMs that Azure reboots together during planned maintenance. Azure guarantees it will only reboot one update domain at a time. Default: 5 UDs per availability set. Azure waits 30 minutes between UDs during maintenance.

Availability Comparison

Proximity Placement Groups

Proximity Placement Groups (PPGs) instruct Azure to co-locate VMs as physically close as possible — in the same datacenter, on the same compute cluster. This minimises network latency between VMs, which is critical for tightly coupled workloads like HPC clusters or multi-tier apps with microsecond latency requirements.

The trade-off: PPGs reduce fault tolerance. If the underlying host cluster or datacenter experiences a failure, all VMs in the PPG may be affected simultaneously. PPGs and full availability zone isolation are inherently in tension — you must choose based on your workload's latency vs. resilience requirements.

Azure Spot VMs

Spot VMs allow you to use Azure's unused compute capacity at up to 90% discount compared to pay-as-you-go pricing. The catch: Azure can evict a spot VM at any time — with only 30 seconds notice — when Azure needs the capacity back for standard-priority workloads.

VM Extensions

VM extensions are small agent packages that run inside the VM to provide post-deployment automation and integration with Azure services. They run under the Azure VM Guest Agent, which must be running on the VM.

VM Snapshots vs. Azure Backup

There are two ways to create point-in-time recovery points for a VM's disks. Understanding when to use each is important for both the exam and production operations.

Azure Hybrid Benefit

Azure Hybrid Benefit (AHB) allows you to apply existing on-premises Windows Server licenses with Software Assurance to Azure VMs — paying Linux pricing for Windows VMs. The savings can reach up to 49% on Windows VM compute costs.

AHB can also be applied to SQL Server licences (avoiding SQL license charges on Azure VMs) and to Azure Stack HCI licensing. For Windows VMs, AHB is toggled at VM creation time or applied to existing VMs without downtime — it is purely a billing flag.

Resizing VMs

VM resize behaviour depends on whether the target size is available on the same physical host cluster as the current VM:

• Same family, nearby sizes: Azure can often resize without deallocating the VM — brief reboot but no full deallocation. Example: D4s_v5 → D8s_v5.
• Cross-family resize: Requires VM deallocation (the VM is stopped and its compute allocation released). The VM is reassigned to a different host cluster that supports the target family. This causes downtime. Example: D4s_v5 → E8s_v5.