Azure Compute Services

AZ-900 AZ-900 · Describe Azure architecture and services

By Abu Anns · Last updated 20 November 2026Exam version: AZ-900 (skills measured as of July 20, 2026)

Three ways to run code

Azure offers three broad ways to run a workload: virtual machines, containers, and functions. The exam expects you to compare them and pick the right one for a situation. The comparison turns on two questions: how much control do you need over the environment, and how much management are you willing to do?

Virtual machines (VMs) are IaaS: a complete virtual computer with its own operating system, running on shared physical hardware. You control the OS, you install what you need, and you manage patching and scaling. VMs suit workloads that need a specific OS, legacy applications, and anything you would previously have run on a physical server. They are the most flexible and the most work.

Containers package an application with its libraries and dependencies into a lightweight, portable unit that runs on a shared operating system. Unlike a VM, a container does not carry its own OS — it shares the host's OS kernel — so containers start in seconds, use fewer resources, and run consistently anywhere: on a developer's laptop, in Azure, or in another cloud. Azure offers Azure Container Instances (run a single container quickly, without managing servers) and Azure Kubernetes Service (AKS) (orchestrate many containers at scale). Containers suit microservices, applications that must run the same way in several environments, and workloads that scale by adding identical copies.

Functions are serverless: you write code that runs in response to an event, and Azure manages everything else. Azure Functions bills per execution and scales to zero when idle. Functions suit short, event-driven tasks — processing an uploaded file, responding to a message, running a scheduled job — and they are the wrong tool for long-running or stateful work.

Virtual machinesContainersFunctions
Service typeIaaSPaaS / IaaS depending on servicePaaS (serverless)
You manageOS, runtime, appApp and its dependenciesCode only
Startup timeMinutesSecondsMilliseconds to seconds
Best forFull control, legacy apps, specific OS needsMicroservices, portability, consistent environmentsEvent-driven, short tasks, scale-to-zero

Virtual machine options

Within VMs, the exam names four options. Azure Virtual Machines are the base offering: pick a size (CPU, memory), an OS image (Windows or Linux), and a region, and you have a server. Azure Virtual Machine Scale Sets (VMSS) create and manage a group of identical, load-balanced VMs that scale automatically with demand — the right choice for a web tier that needs to grow and shrink. Availability sets, covered in the architecture guide, spread VMs across fault and update domains inside one datacenter for a 99.95% SLA. Azure Virtual Desktop is desktop virtualisation: Windows desktops and applications delivered from Azure to any device, with multi-session Windows 10/11 so several users can share one VM — the answer whenever a scenario describes remote workers needing a full Windows desktop from anywhere.

What a virtual machine needs

Creating a VM requires more than the VM itself. The exam lists the supporting resources: a resource group to hold everything; a virtual network for the VM to communicate on; a network interface (NIC) connecting the VM to that network; and storage — a managed disk for the operating system, and optionally data disks. Public IP addresses and network security groups are common additions but the four above are the core requirements. If a question asks what is needed to create a VM, think: group, network, interface, disk.

Application hosting options

When the question is "where should this application run?", the exam offers three hosting answers. Web apps (Azure App Service) host web applications and APIs on a managed platform — deploy code, get scaling, patching, and built-in deployment slots, without managing servers. Containers (Container Instances or AKS) host applications packaged as containers, for portability and microservices. Virtual machines host anything, with full control, at the cost of managing the OS. The decision rule from the service-types guide applies: if the team wants to focus on code, web apps or containers; if it needs the OS, VMs.

Two refinements help with scenario questions. First, App Service plans determine the region, size, and features of the web apps running on them — and multiple web apps can share one plan, which is the cost-efficient way to host several small apps. Second, deployment slots in App Service let you stage a new version (a "staging" slot) and swap it into production with no downtime, rolling back just as easily. Slots are a favourite exam detail for questions about zero-downtime deployments.

Worked example: hosting a new online store

A retailer is building a new online store. The storefront is a web application that must scale for holiday traffic; product images need processing when uploaded; and the stock system is a legacy application requiring a specific Windows Server version. The hosting design: App Service for the storefront (managed platform, autoscaling, deployment slots for safe releases), Azure Functions for image processing (event-driven, scale-to-zero), and a virtual machine for the legacy stock system (full OS control). A scale set could replace App Service if the team wanted VM-level control of the storefront — but since the team wants to focus on code, App Service is the better fit. One application, three compute types, each chosen for its workload.

Choosing compute: a decision checklist

When an exam scenario asks you to pick a compute option, run through these questions in order. First, does the workload need a specific operating system or full control of the environment? If yes, the answer is a virtual machine — nothing else gives you the OS. Second, is the work short and triggered by an event? If yes, the answer is Functions — serverless fits bursts and scheduled tasks better than anything that bills while idle. Third, does the application need to run identically in several environments, or is it built from small independent services? If yes, the answer is containers. Fourth, is it a web application or API the team wants to deploy without managing servers? If yes, the answer is App Service. And finally, does the workload need many identical VMs that scale automatically? If yes, the answer is a scale set. Most scenario questions are decided by the first question that gets a "yes" — work the list from the top and you will land on the right service.

One more distinction the exam probes: scaling up versus scaling out in compute terms. Scaling up a VM means resizing it to a larger size (more CPU and memory) — simple, but limited by the largest available size, and it requires a restart. Scaling out means adding more instances — the scale set and App Service approach — which has no practical ceiling and no downtime, but requires the application to run across several instances. Cloud-native designs favour scaling out; legacy applications that cannot split across instances are stuck scaling up. When a question describes adding instances, think scale out; when it describes a bigger machine, think scale up.

Advertisement

Exam traps to watch for

Practice questions

Test yourself on this guide. Answers are checked instantly, with a full explanation for each question.

Q1.A company runs a web application that must automatically add servers during traffic peaks and remove them afterwards. All servers must be identical and load-balanced. Which Azure service should it use?

Report an error

Q2.A company wants its remote employees to access a full Windows desktop with Office applications from any device, including personal laptops. Which Azure service provides this?

Report an error

Q3.A developer builds an application from small independent services that must run identically on a laptop, in testing, and in Azure, and start within seconds. Which compute approach best fits?

Report an error

Q4.Which resources are required to create an Azure virtual machine?

Report an error

Q5.A team wants to deploy a new version of its web application with no downtime and be able to roll back instantly if something goes wrong. Which App Service feature should it use?

Report an error

Q6.A company runs a batch job that processes uploaded videos. The job runs only when a video arrives, takes a few minutes, and the team wants to pay nothing while no videos are being processed. Which compute option fits best?

Report an error

Q7.What is the main difference between a virtual machine and a container?

Report an error

Q8.A development team wants to host a web application and an API. They want automatic scaling and patching handled by Azure, and they do not want to manage any servers. Which hosting option should they choose?

Report an error