Behavioral Anomaly Detection

Marieke Goethals

DataMiner Behavioral Anomaly Detection will alert you to data sources in your system that suddenly start to show unusual behavior. This could be a value that usually fluctuates and suddenly starts flatlining, a sudden upward or downward spike in a trend, an increase or decrease in the slope of a trend, and so on.

In this use case, we will show some examples of such anomalies: an increase in the temperature of a rack after a server has been removed, and a sudden drop in traffic of an internet exchange platform.

USE CASE DETAILS

As a first example, here is a rack where one server was removed, creating a gap that influenced the airflow.

This is the resulting trend graph. As you can see, DataMiner detected a level shift anomaly because the temperature suddenly rose by about 8 degrees.

A very different example is illustrated here: this trend graph shows the traffic of an internet exchange platform, where a sudden level shift is detected on November 22nd.

The level shift detected by DataMiner corresponds with an actual incident on the platform, which resulted in a drastic drop in traffic.

Want to try out this feature yourself? You can! Go to
docs.dataminer.services.
Our tutorial on detecting anomalies will show you step by step how you can configure your own system to detect anomalies, using example data downloaded from dataminer.services. And if you complete the tutorial, you can even earn some DevOps Points.