AZ-305
Deep Dive
The most repeated question on the AZ-305: where does this workload run? VM, App Service, AKS, Container Apps, Container Instances or Functions. Each has clear signals the exam uses to point you to the answer.
Contents
Each question mirrors exactly the logic the exam uses. Answer in order — the first signal that matches is the answer.
🌿Árbol de decisión — ¿Dónde corre este workload?
Responde las preguntas siguiendo el flujo del examen
Pregunta 1
¿El workload necesita control total del SO, drivers, software legacy, o es lift-and-shift sin cambios en el código?
💡 Señal: "acceso al registro de Windows", "agentes propietarios", "licencias por socket"
| Service | Signals in favor | Signals against |
|---|---|---|
| Virtual Machines | Full control, licensed software/agents, specific GPU, literal lift-and-shift | You manage patches, HA, scale — maximum operational responsibility |
| App Service | HTTP web/API, deployment slots, managed certificates/custom domains, built-in CI/CD | Only HTTP/S (no arbitrary TCP), non-web workloads |
| AKS | Need the Kubernetes API: helm charts, operators, sidecars, multi-cloud portability | Operational overhead of the cluster; if K8s isn't explicitly requested, there's usually a simpler option |
| Container Apps | "Containers without managing Kubernetes", scale-to-zero, events (KEDA), microservices with Dapr | No access to the K8s API — if operators/CRDs are requested, it's AKS |
| Container Instances | Ephemeral/one-off container, virtual nodes for AKS burst, no orchestration | No native autoscaling or load balancing — not for serving apps in production |
| Functions | Event-driven, short executions, "pay only when it runs" (Consumption), bindings to Storage/Cosmos/Service Bus | Long processes (timeout limit on Consumption), complex state (→ Durable Functions) |
| Logic Apps | Integration between systems with connectors (SAP, O365, Salesforce), visual workflows, B2B | Compute-intensive logic — that's code (Functions) |
| Azure Batch | HPC, rendering, simulations — VM pools for massive parallel jobs | Not for serving applications |
Functions hosting — a frequent sub-decision
Consumption: pay per execution, scales to zero, cold starts, limited timeout. Flex Consumption: mitigated cold start + VNet. Premium: no cold start (pre-warmed instances), VNet integration, no practical duration limit. Dedicated (App Service plan): reuse an existing plan. Signal "no cold starts" or "VNet access" → Premium/Flex.
Families (recognizing them is enough)
Cost levers (they drop a lot)
| The scenario says... | Answer |
|---|---|
| Legacy app that requires access to the Windows registry | Virtual Machine |
| REST API with slots for blue/green and managed certificates | App Service |
| Microservices with existing helm charts and operators | AKS |
| Event-driven containers that scale to zero, without managing clusters | Container Apps |
| Containerized job that runs for 2h once a month | Container Instances (or CA Jobs) |
| Process each message arriving in a queue, pay per use | Functions (Consumption) |
| Functions with VNet and no cold start | Functions Premium |
| Orchestrate approval → SAP → email with no code | Logic Apps |
| Video rendering across 500 parallel nodes | Azure Batch (+ Spot) |
| Reduce the cost of stable 24/7 VMs | Reserved Instances + Hybrid Benefit |
| Interruption-tolerant batch at minimum cost | Spot VMs |
Did you get this topic?
Test what you just learned
A team is going to deploy microservices in containers. Requirements: scale to zero outside business hours, scale based on Service Bus queue length, service-to-service communication with built-in discovery, and the team does NOT want to manage Kubernetes clusters. What service do you recommend?
Put what you learned to the test with original AZ-305 exam questions.
Start AZ-305 mock examUnlimited mock exams with the Pro plan
Log in to track your progress.