CaaS

IaaS gives you VMs. Too much to manage for most teams. PaaS takes your code and deploys it, but you lose control over the runtime environment. What if you want something in between? You want to package your app exactly the way you want it (with its dependencies, runtime, and configuration baked in) but not deal with managing servers or Kubernetes clusters.

That is CaaS. Containers as a Service. You build a Docker container image. You hand it to the CaaS platform. It runs it, scales it, monitors it. You do not think about VMs, cluster management, or node provisioning.

AWS ECS with Fargate, Google Cloud Run, and Azure Container Instances are CaaS. You define your container, set resource limits and scaling rules, and the platform handles everything else. It is the sweet spot between PaaS's simplicity and IaaS's control.

Spotify migrated from self-managed Kubernetes to GKE plus Cloud Run for many workloads. Their reasoning: 80% of their services did not need the full power of Kubernetes. CaaS gave those services a simpler deployment path with fewer moving parts, and freed their platform team to focus on the 20% that genuinely needed custom orchestration.

Not all CaaS platforms are the same. The choice depends on your existing cloud provider, pricing model, and whether you need scale-to-zero.