SaveMyCert
Log in
5 of 5 free questions left today·for 10 a day
AZ-305 · Domain 4

Design infrastructure solutions practice questions

Design infrastructure solutions is worth 33% of the AZ-305 exam — the heaviest of the 4 domains. Designing compute (VMs, containers, serverless, batch), application architecture (messaging, events, APIs, caching, configuration, automated deployment), migrations with the Cloud Adoption Framework, and network connectivity, performance, security, load balancing and routing. Official weighting 30–35%. 6 fully worked examples are further down this page, answers included.

Exam weight
33%
the heaviest of the 4 domains
Questions
160
across 4 topics
Free, no account
5/day
a free account allows 10 a day
Explanations
Every option
right and wrong

Build a practice session

5 free questions left today.

Domains

How many?

Mode

Ready when you are

10 fresh questions drawn across 1 of 4 domains, in Learn mode.

Review what you missed

Every question you answer incorrectly, and every question you flag while practising, is saved here automatically. Finish a session and you can come back to re-drill just those.

6 sample Design infrastructure solutions questions, fully explained

Questions from the AZ-305 bank mapped to domain 4, with the answer key and the reasoning behind every option. None of them repeat the examples on the main AZ-305 practice page.

Question 1Design infrastructure solutions

An analytics team is moving an in-memory database to an Azure virtual machine. The workload uses little CPU but needs a high ratio of memory to vCPUs so the whole dataset stays in RAM. Which VM family should you recommend?

Choose one.

  • a
    Memory optimized E-series virtual machines Correct

    E-series sizes are built for a high memory-to-core ratio, which suits in-memory databases and caches.

  • b
    Compute optimized F-series virtual machines

    F-series sizes have a high CPU-to-memory ratio, the opposite of what an in-memory database needs.

  • c
    Burstable B-series virtual machines

    B-series sizes are for low average CPU with short bursts; they don't offer a high memory ratio for a large dataset.

  • d
    Storage optimized L-series virtual machines

    L-series sizes are optimized for local NVMe throughput, not for holding a large dataset in memory.

The concept

VM families are chosen by the workload's resource ratio.

Why that’s the answer

Azure groups VM sizes by their resource shape. Memory optimized sizes (E-series, and M-series for the largest SAP HANA systems) give the most memory per vCPU, which is what an in-memory database needs. Compute optimized F-series favour CPU, B-series are burstable general-purpose sizes, and L-series favour fast local storage.

How to reason it out
  1. Identify the bottleneck resource: memory.
  2. Map it to the family that maximizes memory per vCPU.
  3. Choose memory optimized E-series.

Exam tip: High memory per core: E-series (M-series for the very largest).

AZ-305 Compute Design: Virtual Machines, Scale Sets, Containers, Functions and Azure Batch — the lesson that teaches this.

Question 2Design infrastructure solutions

Fabrikam needs eight jump-box VMs that administrators use only on weekdays between 07:00 and 19:00. Average CPU stays under 10%, with short spikes during sign-in. A jump box must not disappear during those hours, and you must minimize compute costs. Which TWO actions should you recommend? (Select TWO.)

Choose TWO.

  • a
    Deploy the jump boxes on Azure Spot virtual machines

    Spot VMs can be evicted at any time, which breaks the rule that a jump box stays available during working hours, and Spot doesn't offer B-series sizes.

  • b
    Deploy the jump boxes on compute optimized F-series

    F-series sizes are priced for sustained CPU work; an idle jump box would pay for capacity it never uses.

  • c
    Deploy the jump boxes on burstable B-series sizes Correct

    B-series sizes bank credits while CPU is low and spend them on short spikes, which matches a mostly idle machine at the lowest price.

  • d
    Buy a three-year reservation for D-series sizes

    A reservation discounts always-on use of a fixed size; for machines idle most of the time a cheaper burstable size saves more.

  • e
    Schedule auto-shutdown for evenings and weekends Correct

    Shutting down and deallocating the VMs outside working hours stops compute billing while keeping the disks, and admins start them when needed.

The concept

Low-average, spiky CPU fits burstable sizes; idle hours fit scheduled shutdown.

Why that’s the answer

B-series VMs accumulate CPU credits when usage is below their baseline and burst above it when needed, so a mostly idle jump box runs on a cheap size. Auto-shutdown deallocates the VMs outside working hours so compute isn't billed. Spot is cheaper still but can be evicted, which breaks the availability requirement. F-series and reserved D-series both pay for steady capacity the workload doesn't use.

How to reason it out
  1. Read the usage profile: mostly idle with short spikes.
  2. Pick the family priced for that profile: burstable B-series.
  3. Remove hours with no use: auto-shutdown.
  4. Reject Spot because availability is required.

Exam tip: Mostly idle with short spikes: B-series; idle hours: shut down and deallocate.

AZ-305 Compute Design: Virtual Machines, Scale Sets, Containers, Functions and Azure Batch — the lesson that teaches this.

Question 3Design infrastructure solutions

An engineering firm runs a computational fluid dynamics solver that splits one simulation across 32 VMs. The processes exchange data constantly through MPI, and run time is dominated by network latency between nodes. Which VM family should you recommend for the cluster?

Choose one.

  • a
    HB-series high performance compute sizes Correct

    HB-series HPC sizes add an RDMA InfiniBand back-end network with very low latency between nodes, which tightly coupled MPI jobs need.

  • b
    NC-series GPU accelerated compute sizes

    NC-series add NVIDIA GPUs; the solver is CPU-based MPI and its bottleneck is node-to-node latency, not GPU compute.

  • c
    Fsv2-series compute optimized sizes

    Fsv2-series have a high CPU-to-memory ratio but standard Ethernet networking, so they don't fix the latency between MPI processes.

  • d
    Lsv3-series storage optimized sizes

    Lsv3-series have fast local NVMe disks; they do nothing for the inter-node communication the solver waits on.

The concept

Tightly coupled MPI workloads need the RDMA InfiniBand network of HPC sizes.

Why that’s the answer

When one simulation is split across many nodes that exchange messages constantly, latency between nodes dominates. Azure's HPC sizes in the H family, such as HB-series, are the ones with an InfiniBand (RDMA) back-end network. Compute optimized Fsv2 (fast CPUs), GPU NC-series or storage optimized Lsv3 (local NVMe) each improve a different resource but leave the network as the bottleneck.

How to reason it out
  1. Identify the bottleneck: latency between nodes.
  2. Find the sizes with a low-latency RDMA InfiniBand network.
  3. Choose the HB-series HPC sizes.

Exam tip: Tightly coupled MPI across nodes: HPC sizes such as HB-series, with InfiniBand.

AZ-305 Compute Design: Virtual Machines, Scale Sets, Containers, Functions and Azure Batch — the lesson that teaches this.

Question 4Design infrastructure solutions

Contoso spends a steady amount on Azure compute every hour. Over the next three years it plans to move workloads between VM families and regions, and to rebuild some of them on Azure App Service and Azure Container Apps. Contoso wants a commitment discount that keeps applying through those changes without having to manage exchanges. What should you recommend?

Choose one.

  • a
    A three-year Azure savings plan at the shared scope Correct

    A savings plan applies its discount to eligible compute usage across VM families and regions, and to some PaaS compute, so it follows the workloads as they move.

  • b
    Azure Reservations for the current VM sizes in each region

    Reservations discount a specific size family in one region; when workloads move families or regions you must exchange them or lose the benefit.

  • c
    Azure Hybrid Benefit for the Windows Server licences in use

    Hybrid Benefit removes the licence cost for VMs with eligible licences; it isn't a commitment discount and doesn't cover App Service or Container Apps.

  • d
    Azure Spot virtual machines for the steady production workloads

    Spot VMs can be evicted at any time, so they don't suit steady production work, and they aren't a commitment discount.

The concept

A savings plan trades some discount for flexibility across families, regions and services.

Why that’s the answer

Azure Reservations give the deepest discount but are bound to a size family and region (with instance size flexibility inside the group). An Azure savings plan for compute commits to an hourly spend and applies automatically to eligible usage across VM families, regions and selected PaaS compute such as App Service and Container Apps. Because Contoso expects to change families, regions and services, the savings plan keeps paying off without exchanges.

How to reason it out
  1. Note the stable element: steady hourly spend.
  2. Note the moving parts: families, regions, PaaS services.
  3. Pick the commitment that follows usage across all of them: savings plan.

Exam tip: Stable spend, changing shape: savings plan. Stable size and region: reservation.

AZ-305 Compute Design: Virtual Machines, Scale Sets, Containers, Functions and Azure Batch — the lesson that teaches this.

Question 5Design infrastructure solutions

Litware is moving 40 production Windows Server VMs to Azure. It already owns Windows Server Datacenter licences with active Software Assurance and wants to stop paying for the Windows licence again in Azure. What should you recommend?

Choose one.

  • a
    Use Azure Dev/Test pricing for the Windows Server virtual machines

    Dev/Test pricing is limited to non-production use under a Dev/Test offer; these VMs run production workloads.

  • b
    Buy Azure Reservations for the Windows Server virtual machines

    Reservations discount the compute portion; they don't let Litware reuse the licences it already owns.

  • c
    Run the Windows Server virtual machines as Azure Spot instances

    Spot lowers price by allowing eviction, which doesn't suit production VMs and doesn't reuse owned licences.

  • d
    Apply Azure Hybrid Benefit to the Windows Server virtual machines Correct

    Hybrid Benefit lets licences with Software Assurance cover the Windows Server licence, so the VMs bill at the base compute rate.

The concept

Owned licences with Software Assurance are reused through Azure Hybrid Benefit.

Why that’s the answer

Azure Hybrid Benefit applies Windows Server (and SQL Server) licences that have Software Assurance or a qualifying subscription to Azure VMs, removing the licence charge from the VM price. Reservations and Spot reduce the compute charge in other ways, and Dev/Test pricing is only for non-production workloads.

How to reason it out
  1. Identify what Litware wants to stop paying for: the Windows licence.
  2. Find the benefit that reuses owned licences: Azure Hybrid Benefit.

Exam tip: Already own licences with Software Assurance? Turn on Azure Hybrid Benefit.

AZ-305 Compute Design: Virtual Machines, Scale Sets, Containers, Functions and Azure Batch — the lesson that teaches this.

Question 6Design infrastructure solutions

A media company renders video frames each night on a Virtual Machine Scale Set. Each frame is a small task that the render manager retries if a node disappears, and nothing is kept on the VMs. Finance wants the lowest possible cost, including no charges for evicted instances that sit idle. How should you configure the scale set?

Choose one.

  • a
    Use Spot priority with the eviction policy set to Deallocate

    Deallocate stops compute billing but keeps the disks, which still incur storage charges while the instance waits.

  • b
    Use Spot priority with the eviction policy set to Delete Correct

    Spot gives the deepest discount, and the Delete policy removes the VM and its disks on eviction, so nothing keeps billing.

  • c
    Use Regular priority and buy a one-year reservation for it

    A reservation discounts always-on capacity; a nightly render farm would pay for idle reserved hours and gets a smaller discount than Spot.

  • d
    Use Regular priority and enable autoscale on queue length

    Autoscale trims idle capacity, but every instance still bills at the full pay-as-you-go rate rather than the Spot rate.

The concept

Spot fits interruptible work; the eviction policy decides what keeps billing.

Why that’s the answer

The frames are retried automatically and no state lives on the VMs, so eviction is harmless and Spot pricing is the cheapest option. Spot VMs support two eviction policies: Deallocate keeps the VM and its managed disks (disks continue to bill and the VM counts against quota) so it can be restarted, while Delete removes the VM and its disks. For stateless, retried work with a no-idle-charges requirement, Delete is the match.

How to reason it out
  1. Confirm the work tolerates interruption: retried tasks, no local state.
  2. Choose Spot priority for the lowest compute price.
  3. Compare eviction policies: Deallocate keeps billable disks, Delete removes them.
  4. Choose Delete.

Exam tip: Stateless Spot work: eviction policy Delete; need to restart the same VM: Deallocate.

AZ-305 Compute Design: Virtual Machines, Scale Sets, Containers, Functions and Azure Batch — the lesson that teaches this.

What AZ-305 domain 4 tests, topic by topic

The official exam guide breaks Design infrastructure solutions into 4 topics. The question bank follows the same split, so a weak topic shows up as a cluster of misses you can go back and read.

Published AZ-305 practice questions per topic in Design infrastructure solutions
TopicWhat it coversQuestions
Design compute solutionsSkills outline section (AZ-305, skills measured as of April 17, 2026): Specify components of a compute solution based on workload requirements; Recommend a virtual machine-based solution; Recommend a container-based solution; Recommend a serverless-based solution; Recommend a compute solution for batch processing.40
Design an application architectureSkills outline section (AZ-305, skills measured as of April 17, 2026): Recommend a messaging architecture; Recommend an event-driven architecture; Recommend a solution for API integration; Recommend a caching solution for applications; Recommend an application configuration management solution; Recommend an automated deployment solution for applications.40
Design migrationsSkills outline section (AZ-305, skills measured as of April 17, 2026): Evaluate a migration solution that leverages the Microsoft Cloud Adoption Framework for Azure; Evaluate on-premises servers, data, and applications for migration; Recommend a solution for migrating workloads to infrastructure as a service (IaaS) and platform as a service (PaaS); Recommend a solution for migrating databases; Recommend a solution for migrating unstructured data.40
Design network solutionsSkills outline section (AZ-305, skills measured as of April 17, 2026): Recommend a connectivity solution that connects Azure resources to the internet; Recommend a connectivity solution that connects Azure resources to on-premises networks; Recommend a solution to optimize network performance; Recommend a solution to optimize network security; Recommend a load-balancing and routing solution.40
Total160

Revise Design infrastructure solutions before you drill it

AZ-305 Compute Design: Virtual Machines, Scale Sets, Containers, Functions and Azure BatchQuestions on this topic are design scenarios: an organisation describes a workload and two or three requirements, such as no operating system to patch, no container images to build, no Kubernetes skills, scale to zero, Windows-only dependencies, private network access, staged releases, a hard deadline, no shared hardware or the lowest cost, and asks what you should recommend. The options are real Azure compute services, VM families, pricing options, orchestration modes, eviction policies, hosting plans or Batch settings that each miss one requirement, so read every constraint before choosing. Expect near-twin comparisons (Container Apps versus AKS versus Container Instances, Flex Consumption versus Premium versus Dedicated, Flexible versus Uniform scale sets, Deallocate versus Delete, reservations versus savings plans, Dedicated Host versus confidential VMs, Logic Apps Consumption versus Standard, Batch versus CycleCloud) and select-TWO items where each correct option is part of the solution. Current Microsoft guidance matters: Flex Consumption for new serverless function apps, and Flexible orchestration for new scale sets.AZ-305 Application Architecture: Messaging, Events, APIs, Caching, Configuration and DeploymentQuestions on this topic are design scenarios: an organisation describes an application flow and two or three requirements about delivery, ordering, durability, replay, latency, protocols, throughput, regional resilience, consumer limits, where a gateway runs, data freshness, environment isolation, credentials or release risk, and asks what you should recommend. The options are real Azure services, tiers, policies, features or patterns that each miss one requirement, so read every constraint before choosing. Expect near-twin comparisons (Service Bus versus Event Grid versus Event Hubs versus Queue Storage, sessions versus partitioning versus duplicate detection, Geo-Replication versus Geo-Disaster Recovery, the API Management classic and v2 tiers, rate-limit versus quota, revisions versus versions, inbound versus outbound policies, labels versus separate stores, template specs versus deployment stacks, what-if versus validate) and select-TWO items where each correct option is part of the solution. Current Microsoft guidance matters: Azure Managed Redis for new caches, deployment stacks and template specs rather than Azure Blueprints, and workload identity federation rather than stored secrets.AZ-305 Migration Design: Cloud Adoption Framework, Azure Migrate, Database and Data TransferQuestions on this topic are migration design scenarios: an organisation describes its estate (VMware, Hyper-V, physical or other-cloud servers, SQL Server, PostgreSQL, MySQL, Oracle or Db2 databases, file shares or S3 buckets) and two or three constraints, such as a datacenter deadline, no software in guest operating systems, an air-gapped network, a source version, the allowed downtime, the available bandwidth, a need for failback or centrally tracked jobs, and asks what you should recommend. Options are real CAF strategies, Azure Migrate settings, replication methods and migration tools that each miss one constraint. Expect near-twin comparisons (rehost versus replatform, refactor versus rearchitect versus rebuild, replace versus rebuild, appliance versus collector versus CSV import, performance-based versus as on-premises sizing, Managed Instance link versus Log Replay Service versus DMS, Storage Mover versus AzCopy versus Azure File Sync versus Data Box) and select-TWO items that combine steps, such as a Data Box seed with an online catch-up or MySQL Import with data-in replication. Current status matters: Data Migration Assistant is retired, and agent-based dependency analysis no longer onboards new servers.AZ-305 Network Design: Hybrid Connectivity, Private Endpoints, Firewalls and Load BalancingQuestions on this topic are design scenarios: an organisation describes its network, hybrid links or applications and two or three requirements, such as no public IPs on VMs, partners allowlisting source addresses, no traffic over the internet, layer-2 or in-transit encryption, surviving a peering-location outage, MFA through Conditional Access, rules teams can't override, data residency, private-only origins or the least administrative effort, and asks what you should recommend or which settings to choose. Options are real Azure services and tiers that each miss one requirement, so near-twin comparisons dominate: NAT gateway versus outbound rules versus instance public IPs, Bastion Basic versus Standard versus Premium, VPN versus ExpressRoute versus ExpressRoute Direct, MACsec versus IPsec over private peering, private versus service endpoints, Private Resolver inbound versus outbound endpoints, Azure Firewall Standard versus Premium, DDoS IP versus Network Protection, Front Door Standard versus Premium, and Front Door versus Traffic Manager versus Application Gateway versus Load Balancer. Expect troubleshooting-style scenarios where a route, DNS link or peering setting is missing, and select-TWO items where each correct option is part of the solution.

Other AZ-305 domains

Domain 1: Design identity, governance, and monitoring solutions28% of the exam · 135 questionsDomain 2: Design data storage solutions22% of the exam · 105 questionsDomain 3: Design business continuity solutions17% of the exam · 82 questionsAll AZ-305 practice questionsMixed sessions across all 4 domains — the way the real exam interleaves them.

Design infrastructure solutions: your questions

Design infrastructure solutions is domain 4 of the AZ-305 exam guide and carries 33% of the scored content — the heaviest of the 4 domains. On a 50-question paper that works out to roughly 17 questions, though Microsoft Azure does not publish an exact per-domain count and individual exam forms vary.

Source

The domain weight and topic list on this page come from the official AZ-305 exam guide.