PromCon EU 2022: Native Histograms in Prometheus

preview_player
Показать описание
Speaker: Ganesh Vernekar

Histograms in Prometheus have worked reliably for years, but they have had a few downsides when it came to storage efficiency, accuracy of histogram queries, and flexibility in using histograms.

Making histograms efficient and easy to use is a highly complex problem, and given the position of Prometheus we needed to get it right. After years of research by Björn Rabenstein, we have designed a new native histogram that is supported starting with Prometheus v2.40.0.

This talk is accompanied by a brief introduction to native histograms and how they are better. We will see how to start using this and migrate from the old histograms. The talk will end with what to expect in v2.40.0, current limitations, and the future roadmap for native histograms.

Рекомендации по теме
Комментарии
Автор

Congrats on getting out such a big feature! The live release was a nice touch. :)

patricepeterson
Автор

5:59 can someone make me understand how we arrive at those values for positivespans and positivebuckets?

cccc
Автор

8:26 what if we have to move to higher schema values(ie higher resolution)...how do we split bucket entries?

cccc