Rails Orbiting the cloud: Mastering the scalability with Kubernetes
summary
Claims Kubernetes scales Rails from millions to billions of requests per minute and handles 4–5x traffic spikes by combining cluster autoscalers (cluster-autoscaler or Carpenter), KEDA event-driven autoscaling, horizontal pod autoscaling, and tuned node groups. Uses Prometheus/OpenTelemetry for observability, Redis/Memcached for caching, optimized images and podman vs containerd choices for runtime performance. Websockets run on AnyCable in a separate service and succeeds at millions of users. Notes Basecamp moving off cloud with mrsk/KL as a counterexample.
AI-written summary · about
chapters (11)
- 0:00Introduction to talk
- 1:21Rails deployment evolution
- 4:58Why Kubernetes for Rails
- 7:11Measuring scalability in Rails
- 10:03Challenges in scaling Rails
- 13:17Kubernetes reality for Rails
- 15:28Key Kubernetes components
- 19:01Auto scaling strategies
- 21:08Deployment strategies
- 22:53Advanced Rails scaling tips
- 25:49Roadmap for scalable Rails
related talks
01h 03mThe Trail to Scale Without Fail: Rails?
32mOperating Rails in Kubernetes
37m5 Years of Rails Scaling to 80k RPS
28mSimplifying at Scale: 7 Years of Rails Architecture
33mKeynote: Rails Doesn't Scale
19mBuilding an On-Premise Kubernetes...
34mEmbracing Upstream Kubernetes in Web Scale Organization
32mScaling Kubernetes Application to 1M Users