Table of Contents

Proactive cap detection

This DataMiner Analytics feature monitors the global trend in the real-time incoming data of a metric and will automatically generate a notification when the metric is predicted to reach a (critical) low or high value when the trend is extended into the future. To give an accurate estimate of the timing of future issues, proactive cap detection uses a prediction model based on trend data stored in the cloud with STaaS or in a self-hosted Cassandra-compatible database.

Tip

A trend prediction can also be visualized in a trend graph. See Working with trend predictions.

Specifications

For the best results, both real-time and average trending should be activated on a parameter for which you want proactive cap detection to be available. To calculate its predictions, DataMiner Analytics will make use of the available real-time and average data. It can predict at most 200 data points into the future. This is further limited by the available data: if there is a data set of a specific number of points, DataMiner Analytics can never predict further than that number of points divided by ten. For example, if the DataMiner storage contains one year of hourly averages and no daily averages, then DataMiner Analytics computes 365 daily averages and is able to predict issues 36 days into the future.

This feature is currently only available for trended parameters with numeric values, and not for partial table parameters. Because of memory constraints, proactive cap detection is also only possible for up to 100 000 parameters per DMA. If there are more parameters for which proactive cap detection would be possible, no predictions will be available for these and the Analytics log file will mention that the number of tracked parameters exceeded the maximum.

In addition, proactive cap detection is currently only supported for parameters for which there are explicitly specified value bounds. It will predict when a parameter will cross one of these bounds:

  • A high and/or low data range value specified in the protocol, or,

  • A (by default) critical alarm limit of type normal (i.e. not rate or baseline) specified in the alarm template, or,

  • From DataMiner 10.3.11/10.4.0 onwards: A (by default) critical alarm limit of type "absolute" or "relative" specified in the alarm template if either a fixed baseline value is set or a dynamically updated baseline value is configured in the alarm template to detect a continuos degradation, or,

  • A data range indirectly derived from the protocol info. Currently this is limited to the values 0 and 100 for percentage data for which no historical values were encountered outside the [0,100] interval.

However, note that in case there is both a data range in the protocol and an alarm threshold in an alarm template, the alarm template will get precedence.

Proactive cap detection configuration in System Center

In DataMiner Cube, you can enable this feature in System Center, via System settings > analytics config > proactive cap detection. The following settings are available there:

  • Enabled: Allows you to activate or deactivate this feature.

  • Minimum alarm severity: Allows you to configure the lowest alarm threshold severity that will be taken into account for proactive cap detection. If this is for example set to Major, proactive cap detection will alert the operator whenever a parameter is predicted to go out of range or is predicted to trigger a major or critical alarm.

Note

To have proactive cap detection generate automatic notifications for predicted critical low or critical high values, the trend prediction feature also needs to be enabled in System Center.

Suggestion events

The notifications generated by the proactive cap detection feature are displayed in the suggestion events tab of the Alarm Console, along with the notifications for behavioral anomaly detection (see Adding and removing alarm tabs in the Alarm Console) and pattern matching (see Monitoring of trend patterns). These are alarms with severity "Information" and source "Suggestion Engine”.

The value of the suggestion event mentions what kind of issue is expected, e.g. predicted breaches of critical alarm thresholds. The value also mentions "Predicted Critical High" or "Predicted Critical Low".

From DataMiner Cube 10.3.7/10.4.0 onwards, when a trend is projected to reach the minimum or maximum value within a parameter's range, the value of the suggestion event mentions "Predicted maximum value" or "Predicted minimum value", followed by the actual maximum or minimum value of the parameter. The corresponding unit of measurement, as specified in the protocol, will also be included.

Prior to DataMiner Cube 10.3.7/10.4.0, the value of the suggestion event mentions "Predicted above range violation" or "Predicted below range violation".

Propose changes