Gnocchi seems to have rate aggregation. For example, my Packstack cloud configures archive policies ceilometer-low-rate and ceilometer-high-rate that include rate:mean aggregation methods.

Declare such a rate policy in the pipeline and see what you can do with it.

See also https://gnocchi.xyz/releasenotes/4.1....:

New aggregation methods are available for archive policy; rate:mean, rate:last, ....

and https://gnocchi.xyz/stable_4.2/rest.html:

... Those can be prefix by rate: to compute the rate of change ...

I will need this as well at some point in the future; best to start experimenting with this new Gnocchi feature.

Thanks for your help! I used gnocchi "dynamic aggregates" to query and the Operations was: (/ (* 100(metric cpu rate:mean)) (* (* (** 10 9) (metric vcpus mean)) 60) )

but the data point of (metric vcpus mean) was per hour , and the data point of (metric cpu rate:mean) was per min.

so, it can't aggregate. Must change (metric vcpus mean) to instance vcpus number, like this(if instance has 2 cpus): (/ (* 100(metric cpu rate:mean)) (* (* (** 10 9) 2) 60) )