All visualizationsSystem Design · 26 of 34
Kubernetes — the thermostat for containers
Declare "3 replicas". Watch a pod die at 2am and get replaced by arithmetic — then scale by editing a number.
💡
THE BIG IDEA
Docker starts a container in a second; Kubernetes answers the next question — who runs 500 of them and stays calm when they crash? Its whole idea fits in one image: a thermostat. You declare what should be true ("3 replicas of v1"), the cluster records it, and a control loop compares desired with actual, forever. Crashes, scaling and rolling updates all become the same event: a gap between the two numbers, closed automatically. Tap any block for its side of the story.
Press ▶ Play to watch it run step by step, or use the arrow buttons to go at your own pace.
1
The Docker theatre ended on a question: containers start in a second — but who starts 500 of them, watches them, and replaces the ones that die at 2am? Meet the answer.
Kubernetes’ big idea is not "run containers" — Docker does that. It is DECLARED STATE: you say what should be true; the cluster makes it true and KEEPS it true.
🧑‍💻You: "I want 3 replicas"
🚪API server
handling requestjust addedDOWNunder loadhealthy
1/11
UP NEXT IN SYSTEM DESIGN
Serving an LLM — one request that is really five hundred
Watch a prompt get processed once in parallel, then produce answers one token at a time until the GPU runs out of memory.
Railway Seat Allocation — a berth is an interval, not a seatServing an LLM — one request that is really five hundred