SRE Day 2023: Failing to Autoscale?

Resource inefficiency with K8S autoscaling often begins with improper vertical scaling. Then horizontal scaling compounds these issues, which manifest as high cloud costs as the cluster autoscaler adds instances. In this talk, Erwin Daria will inform people how they can use ML to enable effective autoscaling.

Without autoscaling, most companies recognize they are either wasting a lot of resources or risking performance/reliability issues. There’s no way to effectively set resource requests unless your actual usage is completely flat. A way to solve this is by having knowledgeable people look at it all day to make adjustments, or you can just take the financial hit or the risk of instability.

Alternatively one can use technology like machine learning to solve the issue with high accuracy and little to no effort.