KubeCon + CloudNativeCon Europe 2025

Scaling Prometheus to support 50 billion active time series across 20 regions on Kubernetes is a monumental challenge. This session delves into the architecture, processes, and tools that make it possible. We will explore the design of stateful sets and zone-aware deployments to ensure reliability and scalability, alongside deployment processes tailored for high availability and fault tolerance. Learn how cellular architecture enables granular scaling and fault isolation, and discover our approach to multi-tenancy, including protection mechanisms against noisy neighbors such as shuffle sharding, throttling with token buckets . We’ll also discuss the journey of scaling each cell to 1 billion active time series, highlighting the Kubernetes challenges we faced and solved along the way. Attendees will leave with actionable insights into building resilient, efficient, and scalable systems using Kubernetes in the cloud-native ecosystem.