Autoscaling at Scale - How We Manage Capacity at Zalando

Explore autoscaling strategies for managing capacity in large-scale Kubernetes environments through this insightful conference talk. Dive into Zalando's experience with managing approximately 140 Kubernetes clusters of varying sizes. Learn about implementing Horizontal Pod Autoscaler (HPA) and Vertical Pod Autoscaler (VPA) for both stateful and stateless applications, as well as Cluster Autoscaler for efficient resource allocation. Discover limitations, workarounds, and monitoring techniques to ensure sufficient capacity for applications. Gain valuable insights into common pitfalls when scaling with these controllers and explore proposed enhancements for more reliable and efficient autoscaling. Understand the intricacies of infrastructure management, including HPA-Scaling Behavior, multi-container pods, node pools, and control plane autoscaling in a cloud-native environment.

Intro
Infrastructure
Horizontal Pod Autoscaler (HPA)
HPA - Kube Metrics Adapter
HPA-Scaling Behavior
HPA - Multi Container Pods
Cluster Autoscaler (CA) - Node Pools
Cluster Autoscaler (CA) - Fork
Cluster Autoscaler (CA). Node Pool Priority
Control Plane Autoscaler
Open Source