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].
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...