How to Take Prometheus Planet Scale - Massively Large Scale Metrics Installations

Explore how eBay scaled its observability infrastructure to handle massive volumes of time series data in this conference talk from SREcon23 Americas. Learn about the evolution of eBay's metrics installation from a 2 million per second ingest rate in 2017 to 40 million per second with nearly 3 billion active time series. Discover the challenges faced with large centralized clusters and degrading query latencies as cardinality grew. Examine how insights from Google's Monarch paper inspired a decentralized, planet-scale approach using Prometheus TSDB. Delve into the development of field hint indices for efficient query fanout, functional query push down techniques, and the implementation of GitOps for managing globally distributed deployments. Gain valuable lessons on scaling Prometheus to handle planetary-scale metrics installations and the practical challenges of operating such systems.

SREcon23 Americas - How To Take Prometheus Planet Scale: Massively Large Scale Metrics Installations