Overview
Explore the significant memory optimization efforts in Prometheus, the leading CNCF monitoring project, in this 29-minute conference talk by Bryan Boreham from Grafana Labs. Discover how Prometheus' memory usage was reduced by up to 50% over the past year, addressing challenges faced when monitoring millions of metrics. Gain insights into the types of data Prometheus holds in memory, the role of profiling in identifying potential savings, and the impact of transaction isolation on large-scale deployments. Learn about the development of a new data structure for time series labels and understand the implications of these improvements on related projects like Cortex and Thanos. Enhance your knowledge of cloud-native monitoring solutions and their performance optimization techniques.
Syllabus
How Prometheus Halved Its Memory Usage - Bryan Boreham, Grafana Labs
Taught by
CNCF [Cloud Native Computing Foundation]