I wonder how it compares with InfluxData's telegraf. It seem to gather a subset of the metrics collected by telegraf, but maybe it requires far less resources or is meant for scenarios where very frequent measurements are needed.
I've used telegraf to sample at 100Hz for ML workloads. I think one difference is that telegraf pushes each measurement to the DB, while Rezolus summarizes this high granularity data to reduce the amount that needs to be stored.
The most interesting thing is that Rezolus is written in Rust, we may embed it in our project TiKV and let TiKV can trace itself, no need to through outside tools like perf/bcc.
> according to the Nyquist-Shannon Sampling Theorem, the sampling rate must be at least twice the duration of the shortest burst in order to accurately reflect the intensity of a burst.
This statement is a bit sloppy. I think what the author is trying to say is the sampling rate must be such that it samples the shortest burst at least twice. Equivalently you could say that the sampling period should be no longer than half the length of the shortest burst.
This would accurately reflect the fundamental frequency of the burst, but if the burst is not sinusoidal you'll get aliasing. The effect of which will be to falsely indicate lower-frequency bursts that aren't actually present.
7 comments
[ 2.8 ms ] story [ 32.3 ms ] threadThe most interesting thing is that Rezolus is written in Rust, we may embed it in our project TiKV and let TiKV can trace itself, no need to through outside tools like perf/bcc.
It's a bit scary that they consider it OK to spend 15% CPU on all of their servers for telemetry!
This statement is a bit sloppy. I think what the author is trying to say is the sampling rate must be such that it samples the shortest burst at least twice. Equivalently you could say that the sampling period should be no longer than half the length of the shortest burst.
This would accurately reflect the fundamental frequency of the burst, but if the burst is not sinusoidal you'll get aliasing. The effect of which will be to falsely indicate lower-frequency bursts that aren't actually present.