kpi(3.3)

The kpi objects describe how metrics are linked to dimensions and threshold levels.

You may configure kpi objects to perform additional calculations based on metrics, e g ratios or sums of ratios.

The following JSON schema describes the data format of the kpi object type:

Click here to expand...

{
  "$schema": "http://json-schema.org/draft-04/schema#",
  "title": "KpiMgmt Kpi Schema",
  "type": "object",
  "minProperties": 1,
  "additionalProperties": {
    "$ref": "#/definitions/kpi"
  },
  "definitions": {
    "kpi": {
      "type": "object",
      "properties": {
        "node": {
          "type": "array",
          "minItems": 2,
          "items": {
            "type": "string"
          }
        },
        "immediateAlarm": {
          "type": "boolean"
        },
        "windowSize": {
          "type": "integer"
        },
        "threshold": {
          "type": "string"
        },
        "expr": {
          "$ref": "model_expression_schema.json"
        }
      },
      "additionalProperties": false,
      "required": [
        "node",
        "windowSize",
        "expr"
      ]
    }
  }
}
{
  "$schema": "http://json-schema.org/draft-04/schema#",
  "title": "KpiMgmt Expression Schema",
  "type": "string"
}

The kpi object type has the following properties:

Property	Description
node	`node` must contain an array of strings that constitutes a path to a node in a `tree` object. The first value in the array must match the name of a tree object, and the subsequent values must match the name of the descendant nodes in the `tree` object.
immediateAlarm	When `immediateAlarm` is set to `false`, exceeded threshold levels will be indicated in a `KPIOutput` UDR at the end of a period. The UDR field `outputType` will be set to 0. This is the default behavior. When `immediateAlarm` is set to true, a `KPIOutput` UDR with `outputType` 1 will be generated at the time of detection. Exceeded threshold levels will also be indicated in the output that is generated at the end of the period. Immediate alarm data are sent by Spark to the Kafka topic `kpi-alarm`.
windowSize	The incoming data is sorted into time periods based on the timestamp value of the `KDR` UDRs. The KPIs are calculated and generated according to these periods. `windowSize` defines the length of the periods in the same time unit as the KDR timestamp. A flooring algorithm is applied to timestamps in order to select the start time of a period: `period start = kdr.timestamp - (kdr.timestamp mod windowSize)` Note! When the value of `aggregated-output` is `false`, `kpi` objects in your service model may have different window sizes. However, when the value of `aggregated-output` is true, all `kpi` objects in the model must have the same window size.
threshold	`threshold` may contain the name of a `threshold` object.
expr	`expr` must contain an arithmetic expression based on `metric` objects. You can use the expression to calculate e.g. ratios or sums of ratios. Example - Expressions "Plain" metric: `"expr": "AvgLength"` Ratio: `"expr": "TotalAttempts/TotalSuccessful"` Sum of Ratios: `"expr": "(Type1Attempts/Type1Successful)+(Type2Attempts/Type2Successful)"` In the case of division by zero, the value of the output of the expression will be positive infinity, negative infinity, or NaN (Not a Number) as defined in the JVM specification.

Example - JSON Representation

"kpi": {
     "Region.AL": {
       "node": [
         "tree1",
         "Region"
       ],
       "windowSize": 60,
       "threshold": "Region.AL",
       "expr": "AvgLength"
     }
 }