ALERTING based on
ADVANCED ANOMALY DETECTION
Most tools detect simple threshold-based anomalies, making it difficult to distinguish false alarms from real issues. With Wavefront, you create smart alerts that dynamically filter noise and capture true anomalies.
The top chart shows the volume of traffic into a front-end load balancer. There are two distinct features in this chart:
- A vertical dip on the left side of the chart, which is a false alarm
- A much shallower, but more sustained dip towards the right side of the chart, which represents a true outage.
Let’s assume we used threshold-based alerts. Even if we smoothed the line by taking the moving average, you would see a triggered alarm for both dips.
Further, in the bottom chart, we use a function to show the moving median of the top chart. The moving median removes the first short dip, but not the second dip. Therefore, we can create an alert condition in Wavefront based on the moving median suddenly dipping. With hundreds of analytics functions, Wavefront helps you craft the perfect alert for any given anomaly.