Use case 3: Compute cost optimization

Now that we’ve explored how KEDA and Karpenter can help you autoscale web-based applications and job-based workloads, let’s take a closer look at another common use case where Karpenter helps you to cost-optimize your compute costs in a cluster. We’ll focus on what you can do with Karpenter, and how, by only making a few changes to the NodePool (and sometimes to your application), you can reduce the compute costs dramatically.

During the following lab, you’ll start with a non-optimized workload so that you can see the cost reduction while you make the changes. You’re going to use a visualization tool called eks-node-viewer, which shows you the efficiency of a cluster in terms of the allocated resources, the estimated monthly compute cost, and the hourly price, among other things. Then, we’ll introduce optimizations such as using ARM-based instances, right-sizing overprovisioned pods, and using spot...