ACI vs AKS vs Container Apps: When to Use What

Azure provides three container compute services — Azure Container Instances (ACI), Azure Kubernetes Service (AKS), and Azure Container Apps. Choosing the right one depends on your workload characteristics, team expertise, and operational requirements. This lesson provides a structured decision framework, compares the services across multiple dimensions, and walks through real-world scenarios.

---

The Decision Spectrum

Think of the three services as a spectrum from simple to complex:

Simple                                                    Complex
  |                                                          |
  ACI              Container Apps                         AKS
  |                      |                                   |
  Single containers      Microservices                  Full K8s
  Batch jobs             APIs / web apps               Custom operators
  Dev/test               Event-driven                  Multi-team platform
  Short-lived            Scale to zero                 Stateful workloads

---

Comparison Matrix

Dimension	ACI	Container Apps	AKS
Abstraction level	Run a container	Run an app (serverless)	Run a cluster (platform)
Kubernetes knowledge	None required	None required	Required
Scaling	Manual	Automatic (0 to N)	Automatic (HPA, KEDA, cluster autoscaler)
Scale to zero	Stop container group	Yes (built-in)	Yes (with KEDA + cluster autoscaler)
Ingress	Public IP / DNS label	Built-in Envoy + custom domains	You manage (NGINX, App Gateway)
TLS	Manual	Managed or custom	You manage (cert-manager)
Service mesh	No	No (Dapr instead)	Yes (Istio, Linkerd)
Persistent storage	Azure Files, emptyDir	Azure Files	Azure Disk, Azure Files, Blob
Stateful workloads	Limited	Limited	Full support (StatefulSets)
CI/CD integration	Basic	GitHub Actions, Azure DevOps	Full (GitOps, Flux, Argo CD)
Networking	VNet injection	Managed VNet, custom VNet	Full VNet, CNI, network policies
Multi-container	Container groups	Sidecar containers	Pods (any number of containers)
GPU support	Yes	Yes (dedicated plan)	Yes
SLA	99.9%	99.95%	99.95% (Standard), 99.99% (Premium)
Cost model	Per-second (CPU + memory)	Per-second (consumption) or dedicated	Node VM costs

---

Decision Tree

Use this step-by-step guide to choose the right service:

Step 1: What Is the Workload?

If your workload is...	Consider...
A one-off script or batch job	ACI or Container Apps Job
A long-running API or web service	Container Apps or AKS
A scheduled task (cron)	Container Apps Job or AKS CronJob
An event-driven processor (queues, events)	Container Apps or AKS with KEDA
A complex multi-service application	AKS
A machine learning training pipeline	AKS (GPU node pools)
A CI/CD build agent	ACI

Step 2: What Is Your Team's Kubernetes Expertise?

If your team...	Choose...
Has no Kubernetes experience	ACI or Container Apps
Has some Kubernetes knowledge	Container Apps
Has deep Kubernetes expertise	AKS
Wants to learn Kubernetes	Start with Container Apps, graduate to AKS

Step 3: What Operational Model Do You Want?

If you want...	Choose...
Zero infrastructure management	ACI or Container Apps
Managed infrastructure with some control	Container Apps
Full control over the platform	AKS
To run custom Kubernetes operators	AKS
GitOps-based continuous deployment	AKS

---

Real-World Scenarios

Scenario 1: E-Commerce API

Requirements: REST API serving a web frontend, auto-scaling based on traffic, custom domain with TLS, connection to Azure SQL Database.

Best choice: Container Apps

Why:

• Built-in HTTP ingress with custom domains and managed TLS
• Automatic scaling based on HTTP concurrency (including scale to zero off-peak)
• Simple deployment model — no Kubernetes knowledge required
• VNet integration for secure database connectivity

Scenario 2: Video Processing Pipeline