One of the top benefits of Kubernetes is efficiency. Nevertheless, several companies adopting Kubernetes may experience high costs and performance issues. The challenge of manually tuning Kubernetes applications is well known to Performance Engineers and SREs.

During his session at Conf42 Cloud Native 2022, Giovanni Gibilisco (Head of Engineering at Akamas) talks about this challenge and also shows how to overcome it thanks to a new approach that leverages ML techniques.

In the first part of his speech, Giovanni explains some less-known facts about Kuberenetes resource management and auto-scaling mechanisms. He then demonstrates how to get a Kubernetes microservices application automatically tuned for both pod and runtime configurations. The real-world case presented refers to an organization whose optimization goal was to both minimize the Kubernetes cost and maximize the application throughput, while also matching their SLOs.