In this presentation at Performance Summit 2021 on September 29th, Stefano Doni (CTO at Akamas) describes how ML techniques enable Developers and SREs to automatically optimize Kubernetes for cost efficiency and reliability.
Kubernetes the task of properly configuring pod resources (requests and limits) across many microservices can be very complex. This often translates into higher infrastructure costs and may also lead to service availability and performance issues.
During this session, an introduction to Kubernetes resource management concepts is first provided, as a base for understanding these challenges. Then Machine Learning techniques are introduced to demonstrate how they help to automatically identify the size of pod resources that both minimize infrastructure cost and improves application performance and reliability.
See also the other session by Akamas at Performance Summit 2021 here.