Kubernetes Usurped Docker's Orchestration

This was the moment Docker lost the right to charge for the control plane. Orchestration is the layer that decides which machine runs each container, restarts failed workloads, and rolls out updates across a cluster. Docker tried to sell that layer to ops teams with Swarm and Docker Enterprise. Google and the cloud vendors turned that layer into shared infrastructure, open sourced in Kubernetes, standardized through CNCF, and bundled into managed cloud products.

• Kubernetes spread because it was not just free software, it became the default common language across clouds. Google open sourced Kubernetes in 2014, released v1.0 in July 2015, and transferred it into CNCF as the foundation’s first hosted project in March 2016. That made it easier for AWS, Microsoft, IBM, and others to rally around one standard instead of a Docker controlled product.
  2 google 3 googleblog 4 cncf
• The monetization models were different. Docker sold orchestration as enterprise software, which meant long sales cycles into operations teams that often were not the users feeling the daily value. Google and AWS monetized the same need indirectly, by charging for compute, storage, and managed Kubernetes services like Google Container Engine and Amazon EKS.
  1 sacra 5 aboutamazon 6 sacra
• The market outcome showed up in Docker’s restructuring. In November 2019, Docker sold its Enterprise platform business to Mirantis and kept the developer tools and brand. That split reflected where value had really accumulated, developers loved Docker for building and running containers locally, while orchestration had become infrastructure plumbing owned by clouds and Kubernetes vendors.
  1 sacra 6 sacra 7 globenewswire

Going forward, orchestration remains important but it is no longer the scarce product. The scarce layer is developer workflow on top of Kubernetes, local build and test, security checks before code ships, and tools that save time for engineering teams. That is why Docker’s second act moved up the stack, away from selling the cluster and toward serving the developer sitting in front of the laptop.