Containers package an application with its dependencies into a portable, isolated unit that runs the same on a laptop, a CI runner, or a production cluster. Orchestration is what runs many containers across many machines - scheduling, scaling, networking, and recovery. Together they are the standard way data platforms ship and operate: Spark on Kubernetes, Airflow tasks in pods, dbt in a container, and every microservice around them.

The space splits into a few layers. Image build tools (Docker, Buildah, Kaniko) turn a Dockerfile into an image. Runtimes (containerd, CRI-O) actually execute containers. Orchestrators (Kubernetes, Docker Swarm) run them at scale, with Kubernetes the dominant standard. On top of Kubernetes sit distributions and managers (OpenShift, Rancher) and the cloud-managed services (EKS, AKS, GKE) that operate the control plane for you.

For data engineers, the relevance is concrete: most modern orchestration and processing frameworks run on Kubernetes, isolating dependencies and scaling compute to zero between jobs. The pages here cover the tools across all these layers, from building images to running clusters.