topic Re: Performance Capacity calculation raw to final in ONTAP Discussions

Performance Capacity calculation raw to final

JONATHAPQ — Wed, 04 Jun 2025 13:07:45 GMT

I'm trying to work out how "current_utilization" in the -raw output from "statistics show -object resource_headroom_cpu -raw" correlates (ie. is calculated/converted) to main percentage output in "statistics show -object resource_headroom_cpu"

EXAMPLE:

cluster1::*> statistics show -object resource_headroom_cpu

Object: resource_headroom_cpu
Instance: CPU_cluster1-n1
Start-time: 11/9/2018 10:23:20
End-time: 11/9/2018 10:31:39
Elapsed-time: 498s
Scope: cluster1-n1

Counter Value
-------------------------------- --------------------------------
current_latency 330us
current_ops 4215
current_utilization 56% <<<<<<<<<<<<<

...

cluster1::*> statistics show -object resource_headroom_cpu -raw

Object: resource_headroom_cpu
Instance: CPU_cluster1-n1
Start-time: 11/9/2018 10:27:43
End-time: 11/9/2018 10:27:43
Scope: cluster1-n1

Counter Value
-------------------------------- --------------------------------
current_latency 20267702036037us
current_ops 53787499981
current_utilization 36639819423391% <<<<<< (how is this 56%?)

Similarly, how is -raw optimal_point_utilization converted to standard output?

EG.

statistics show -object resource_headroom_cpu:

optimal_point_utilization 88

statistics show -object resource_headroom_cpu -raw:

optimal_point_utilization 28078837

Re: Performance Capacity calculation raw to final

Freedom — Mon, 03 Dec 2018 13:06:02 GMT

Raw headroom counters are made available to OPM for advanced computations.

Raw counter:Means that the counter being exported has no math associated with it.

In order for raw counters to be processed into so called "cooked counters" subsequent to retrieval from the database, certain meta-data for each counter is required.

In your example, RAW counter represent % of utilization ( how many microseconds/second CPU was busy with specific work - percentage of time one or more threads is executing "work" ) and has quite complex math behind it.

Recomendation is to use COOKED counters only.

Re: Performance Capacity calculation raw to final

JONATHAPQ — Mon, 03 Dec 2018 15:30:41 GMT

Thanks for the reply. The reason I'm looking to get at this is because the APIs return raw values and these are what we are using with our 3rd party monitoring platform, so I'm trying to mimic OPM.

Is there not a way of using the API to pull the "cooked" values that you get on CLI, rather than the RAW?