Specify custom upper and lower metric bounds
Use the Bounds Settings wizard to override learned control bounds by specifying custom upper and lower bounds for an Metric Intelligence metric.
Before you begin
About this task
Metric Intelligence calculates upper and lower control bounds for metrics using statistical models learned from historical metric data. Upper and lower bounds are then used in the analysis for detecting anomalous CIs or resources, affecting anomaly scores. Based on historical data, some CIs, resources, or metrics might have values that you specifically want or do not want to be notified about, despite the calculated ranges. For example, trying to meet an SLA without allowing a metric to exceed or be below a certain value for too long without being notified. When concrete upper or lower bound values are known for a metric, you can configure that metric with custom bounds to override the calculated bounds. For example, a CPU metric which does not exceed 85%.
Use the Bounds Settings wizard to specify custom static bounds or ranges of bounds, and to preview anomaly analysis of actual data based on the custom settings. The metric class and the configuration setting rule that custom bounds must be associated with, are both automatically generated by the Bounds Settings wizard.
- Static upper and lower bounds
- Custom static upper and lower bound values are used as the control bounds in effect, completely overriding the learned values for upper and lower bounds for the metric. The median of the metric data must be within the specified static bounds range, otherwise the static bounds are not applied.
- Ranges for upper and lower bounds
- Custom upper and lower bounds, each expressed as a range of values. The
bounds in effect are a combination of the custom minimum and maximum
values that define the bound range, and the learned upper and lower
bounds:
- If a learned bound value for a point in time is within the custom range of upper or lower bound, then the learned bound value is in effect.
- If a learned bound value for a point in time is outside the custom range, then the Max or the Min values of the custom range are in effect respectively. If the learned value exceeds the custom range, then the custom Max value is in effect, and if below then the Min custom value is in effect.
When a metric value is outside a control bound, the deviation from the control bound is normalized by a width value that the system calculates from data. Lower and upper width values determine how much a metric value needs to exceed the control bounds to be anomalous. Changing a width for the control bounds affects the speed of reaching a high anomaly score and being notified about the anomaly. In addition to setting custom control bounds, you can override width values by specifying custom deviation and time values that the system uses to calculate custom widths values.
The system stores custom control bounds and custom widths for a metric in a metric class. Custom values in a metric class override the learned bounds and widths for the metric, and are then used in subsequent anomaly analysis.