Now that the app is running in Kubernetes, how do we scale it to meet demand? What metric should we use? CPU? Requests? something else? Let's dig into why we auto-scale, and how we auto-scale with lots of examples. Finally we'll look at potential pitfalls and gotchas like how to scale to 0 and how to avoid scaling too big for your budget. Come learn how to scale with Kubernetes. It's easy to imagine auto-scaling. "Why isn't it just built in?" It depends a lot on the metric you choose. Is it an I/O-bound system? Then request length or concurrent requests is better. Is it a compute-bound problem? Then measuring CPU or memory usage is a better metric. We'll show examples of Kubernetes autoscalers including an I/O-bound one, a CPU-bound one, and a custom metric. Attendees leave with a GitHub repo containing all the code so they can clone it and use it in their own work. [h2][b][url=https://robrich.org/slides/k8s-autoscale/]Slides[/url][/b][/h2]