Workload Autoscaling and KEDA

By now, you’ve seen what the HPA and VPA can do, and you’ve probably noticed their limitations. Scaling based on CPU or memory works fine for some workloads, but what about when your application needs to scale based on queue depth? Or, when you want to scale to zero during off-hours to save costs? When is latency a better indicator of load than CPU utilization? This is where KEDA comes in.

KEDA extends the built-in HPA without replacing it. Instead of being limited to CPU and memory metrics, KEDA lets you scale based on pretty much any event source you can think of – for example, message queues, databases, HTTP traffic, scheduled times, cloud provider metrics, and more. It works alongside the HPA, giving you the flexibility that your workloads actually need.

This part covers KEDA in depth, starting with the basics of how it works and why so many teams have adopted it. You’ll learn how to configure scaling rules based...