When should you choose Azure App Service vs Container Apps?

Two managed compute options on Azure that overlap heavily. Pick based on your operational needs.

Azure App Service

• PaaS for web apps: deploy .NET, Node, Python with zero ops
• Built-in slot-based deploys, auto-scale up to ~30 instances
• One container or one runtime per app — no sidecars

az webapp up --runtime "DOTNETCORE:9.0" --name prepstack-api

Cost: ~$13/mo (B1, 1.75 GB RAM, always-on) up to ~$280/mo (P1V3, 8 GB RAM, advanced autoscale).

Azure Container Apps

• Managed Kubernetes (you don't see the cluster), built on KEDA + Dapr
• Scale-to-zero when idle
• Multi-container support (sidecars OK)
• Built-in revision-based traffic splitting for canaries
• Event-driven autoscale (HTTP, Kafka, Service Bus)

properties:
  template:
    containers:
      - { name: api, image: ghcr.io/me/api:1.4 }
    scale:
      minReplicas: 0          # ← scale to zero
      maxReplicas: 20
      rules:
        - name: http
          http: { metadata: { concurrentRequests: "30" } }

Cost: pay-per-vCPU-second + memory-second. Idle workloads ≈ free. Active workload pricing similar to App Service Premium but more granular.

Decision matrix:

Need	App Service	Container Apps
Single .NET / Node / Python app	✓ simplest	✓ overkill
Multiple containers (sidecars)	✗	✓
Scale to zero (cost optimisation)	✗	✓
Slot swap deploys	✓ native	✓ via revisions
Custom routing / Dapr	✗	✓
Windows containers	✓	✗
Learning curve	Lowest	Medium

Start with App Service unless you specifically need multi-container, scale-to-zero, or Dapr. Migration App Service → Container Apps is straightforward; the reverse is rarely worth it.