Default monitor settings

When you add an anomaly detection monitor from Explorer without manually editing the monitor, the monitor uses default settings based on the metric type. Note that Compare metrics (Compare aggregates, Row by row) only support manual thresholds, which do not have defaults. Also note that default training periods depend on whether the metric has seasonality.

Aggregation metric

Option

Default setting

Symptom to detect

Value out of Expectations

Feature

Value

Compare against

N/A

Bounds

tolerance interval

Account for trend

Disabled

Forecasting

Disabled

Training period (no seasonality)

Default training periods include the most recent data collection, and extend into the past based on the metric's Aggregation Interval or Polling Interval:

  • If the interval is Hourly or shorter, training period is one week
  • If the interval is Daily, training period is 30 days
  • If the interval is Weekly, training period is 120 days
  • If the interval is Monthly or longer, training period is 15 intervals

Training period (seasonal)

Default training period includes the most recent data collection, and extends 15 season periods into the past.

Drift duration

0s (zero seconds)

Recovery duration

Equal to the Aggregation Interval or Polling Interval

Aggressiveness

1

Drift direction

Both

Incident learning

Enabled

Conformity metrics

Option

Default setting

Symptom to detect

Value out of Expectations

Feature

Value

Compare against

N/A

Bounds

tolerance interval

Account for trend

Disabled

Forecasting

Disabled

Training period (no seasonality)

Default training periods include the most recent data collection, and extend into the past based on the metric's Aggregation Interval or Polling Interval:

  • If the interval is Hourly or shorter, training period is one week
  • If the interval is Daily, training period is 30 days
  • If the interval is Weekly, training period is 120 days
  • If the interval is Monthly or longer, training period is 15 intervals

Training period (seasonal)

Default training period includes the most recent data collection, and extends 15 season periods into the past.

Drift duration

0s (zero seconds)

Recovery duration

Equal to the Aggregation Interval or Polling Interval

Aggressiveness

1

Drift direction

Both

Incident learning

Enabled

Custom SQL metrics

Option

Default setting

Symptom to detect

Value out of Expectations

Feature

Value

Compare against

N/A

Bounds

tolerance interval

Account for trend

Disabled

Forecasting

Disabled

Training period (no seasonality)

Default training periods include the most recent data collection, and extend into the past based on the metric's Aggregation Interval or Polling Interval:

  • If the interval is Hourly or shorter, training period is one week
  • If the interval is Daily, training period is 30 days
  • If the interval is Weekly, training period is 120 days
  • If the interval is Monthly or longer, training period is 15 intervals

Training period (seasonal)

Default training period includes the most recent data collection, and extends 15 season periods into the past.

Drift duration

0s (zero seconds)

Recovery duration

Equal to the Aggregation Interval or Polling Interval

Aggressiveness

1

Drift direction

Both

Incident learning

Enabled

Data delay metrics

OptionDefault setting
Symptom to detectValue out of Expectations
FeatureValue
Compare againstN/A
Boundstolerance interval
Account for trendDisabled
ForecastingDisabled
Training period (no seasonality)Default training period includes the data collection window one week before the most recent window.
Training period (seasonal)Default training period includes the most recent data collection, and extends 15 season periods into the past.
Drift duration0s (zero seconds)
Recovery duration1h (one hour)
Aggressiveness5
Drift directionUp
Incident learningEnabled

Data volume metrics

Option

Default setting

Symptom to detect

Value out of Expectations

Feature

Percentage change

Compare against

Last sample

Bounds

tolerance interval

Account for trend

Enabled

Forecasting

Disabled

Training period (no seasonality)

Default training periods include the most recent data collection, and extend into the past based on the metric's Aggregation Interval or Polling Interval:

  • If the interval is Hourly or shorter, training period is one week
  • If the interval is Daily, training period is 30 days
  • If the interval is Weekly, training period is 120 days
  • If the interval is Monthly or longer, training period is 15 intervals

Training period (seasonal)

Default training period includes the most recent data collection, and extends 15 season periods plus one aggregation/polling interval into the past.

Drift duration

0s (zero seconds)

Recovery duration

Equal to the Aggregation Interval or Polling Interval

Aggressiveness

7

Drift direction

Down

Incident learning

Enabled

Distribution metrics

Option

Default setting

Symptom to detect

Value out of Expectations

Feature

Value

Compare against

N/A

Bounds

tolerance interval

Account for trend

Disabled

Forecasting

Disabled

Training period (no seasonality)

Default training periods include the most recent data collection, and extend into the past based on the metric's Aggregation Interval or Polling Interval:

  • If the interval is Hourly or shorter, training period is one week
  • If the interval is Daily, training period is 30 days
  • If the interval is Weekly, training period is 120 days
  • If the interval is Monthly or longer, training period is 15 intervals

Training period (seasonal)

Default training period includes the most recent data collection, and extends 15 season periods into the past.

Drift duration

0s (zero seconds)

Recovery duration

Equal to the Aggregation Interval or Polling Interval

Aggressiveness

1

Drift direction

Both

Incident learning

Enabled

Null percent metrics

Option

Default setting

Symptom to detect

Value out of Expectations

Feature

Value

Compare against

N/A

Bounds

tolerance interval

Account for trend

Disabled

Forecasting

Disabled

Training period (no seasonality)

Default training periods include the most recent data collection, and extend into the past based on the metric's Aggregation Interval or Polling Interval:

  • If the interval is Hourly or shorter, training period is one week
  • If the interval is Daily, training period is 30 days
  • If the interval is Weekly, training period is 120 days
  • If the interval is Monthly or longer, training period is 15 intervals

Training period (seasonal)

Default training period includes the most recent data collection, and extends 15 season periods into the past.

Drift duration

0s (zero seconds)

Recovery duration

Equal to the Aggregation Interval or Polling Interval

Aggressiveness

1

Drift direction

Both

Incident learning

Enabled

Row count/Byte count metrics

Option

Default setting

Symptom to detect

Value out of Expectations

Feature

Percentage change

Compare against

N/A

Bounds

tolerance interval

Account for trend

Enabled

Forecasting

Disabled

Training period (no seasonality)

Default training periods include the most recent data collection, and extend into the past based on the metric's Aggregation Interval:

  • If Aggregation Interval is Hourly or shorter, training period is one week plus one aggregation interval
  • If Aggregation Interval is Daily, training period is 31 days
  • If Aggregation Interval is Weekly, training period is 127 days
  • If Aggregation Interval is Monthly or longer, training period is 16 aggregation intervals

Training period (seasonal)

Default training period includes the most recent data collection, and extends 15 season periods plus one aggregation interval into the past.

Drift duration

0s (zero seconds)

Recovery duration

Equal to the Aggregation Interval

Aggressiveness

7

Drift direction

Down

Incident learning

Enabled

Update delay metrics

Option

Default setting

Symptom to detect

Value out of Expectations

Feature

Value

Compare against

N/A

Bounds

tolerance interval

Account for trend

Disabled

Forecasting

Disabled

Training period (no seasonality)

Default training periods include the most recent data collection, and extend into the past based on the metric's Aggregation Interval:

  • If Aggregation Interval is Hourly or shorter, training period is one week
  • If Aggregation Interval is Daily, training period is 30 days
  • If Aggregation Interval is Weekly, training period is 120 days
  • If Aggregation Interval is Monthly or longer, training period is 15 * Aggregation Interval

Training period (seasonal)

Default training period includes the most recent data collection, and extends 15 season periods into the past.

Drift duration

0s (zero seconds)

Recovery duration

Equal to the Aggregation Interval

Aggressiveness

5

Drift direction

Up

Incident learning

Enabled