Cortex is a scalable, highly available, multi-tenant, long term storage for Prometheus. It is designed to reliably ingest and query millions of time series per second, while keeping the operational cost low. But if you try to get started, it is easy to get lost in the tens of different components and config options. In this talk, we will introduce Cortex and its architecture in an approachable way. We’ll start from the minimum setup required and we’ll iteratively cover the Cortex core design principles, sharding, replication, failure modes and a series of techniques employed by Cortex for accelerating PromQL queries. Finally, we’ll look at what the future holds for us and the upcoming features and improvements in Cortex. Cortex is an open source project, released under the Apache 2.0 license, and in the CNCF sandbox stage.


Please login to leave a comment

Matteo Vivona at 11:29 on 22 Oct 2020

One of the best talk so far

Really interesting glimpses of an architecture of a production-grade distributed application.

Anonymous at 11:49 on 22 Oct 2020

Marco knows his stuff.

Alberto at 11:54 on 22 Oct 2020

Great high level presentation and awesome deepdive

Really great presentation on Cortex and its under-the-hood architecture.

Great presentation

practical and informational, deep but not *too* much, very good!

very interesting, I didn't know cortex

M_a_s_s_i at 06:37 on 26 Oct 2020

A handy insight into Prometheus storage internals! Thanks :-)