Skip to content

Comment on Big Prometheus: Thanos, Cortex, M3DB and VictoriaMetrics at Scale

Comments

VictoriaMetrics author here. I like the post, since it is cleanly written and it isn't biased to certain solution. I'd recommend readers trying all the mentioned solutions - Thanos, Cortex, M3DB and VictoriaMetrics and then choosing the solution that fits them the best.

Each solution has its own weak and strong points. The main selling points for VictoriaMetrics are:

* Operation simplicity. This is especially true for a single-node version, which is represented by a single self-contained binary without any dependencies. It is configured by a few command-line flags, while the rest of configs have sane defaults, so they shouldn't be touched in most cases.

* Low resource usage (CPU, RAM, disk space and iops, network bandwidth).

* High performance.

See also an interesting talk from PromCon 2019, where all these solutions are compared by Adidas monitoring team [1].

[1] https://promcon.io/2019-munich/talks/remote-write-storage-wa...

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.