performance counter & rest api

[jongyun]

Hi Experts.

I have some mathmetical problem using statistic to monitor performance.
The below is result of statistics API (# /api/cluster  with ontap 9.6)
Large digits of number has shown and I would like to change this into milli second measurement.
For the information, statistics API poll data every 3 seconds, which is near real-time.
Any idea or comments on this?

and then what is latency unit of metric.*

what are the 392units below (ms or us)

"latency": {

      "other": 2060,

      "total": 392,

      "read": 469,

      "write": 237

Please provide a detailed description of metrics.* and stastics.* 

####output of /api/cluster###########

"metric": {

    "timestamp": "2020-07-14T07:30:30Z",

    "duration": "PT15S",

    "status": "ok",

    "latency": {

      "other": 2060,

      "total": 392,

      "read": 469,

      "write": 237

    },

    "iops": {

      "read": 23822,

      "write": 22332,

      "other": 970,

      "total": 47125

    },

    "throughput": {

      "read": 2648275729,

      "write": 365964117,

      "other": 584936,

      "total": 3014824783

    }

  },

"statistics": {
    "timestamp": "2020-07-14T07:31:00Z",
    "status": "ok",
    "latency_raw": {
      "other": 8052914094221,
      "total": 79593320562765,
      "read": 50965207630433,
      "write": 20575198838111
    },
}

[loudspeakers]

I'm having the same question.

What I know is that in metrics, there is an average value for last 15s and under statistics, it's sum of values from boot (counter that only goes up).

Looking at statistics volume show command, the latency is reported in us, so that might be the case here too... 

[loudspeakers]

I read through REST API docs and this is the explanation for cluster statistics:

statistics: These are raw performance numbers, such as IOPS latency and throughput. These numbers are aggregated across all nodes in the cluster and increase with the uptime of the cluster.

latency_raw description: The raw latency in microseconds observed at the storage object. This can be divided by the raw IOPS value to calculate the average latency per I/O operation.

[jcolonfzenpr]

I have better results with this kind of questions asking on the netapp slack channel.

https://join.slack.com/t/netapppub/shared_invite/zt-ki0sse86-6ihXPApFepvu0Nx~YibCtA