AI Engine

AI Baselines

Know what normal looks like for every device

A single fixed threshold rarely fits every device. A core switch that runs at 70% on a Monday morning is healthy; an access switch at the same level at midnight may not be.

AI Baselines learn the normal behavior of each monitored metric from the history your own SoldierX AI NMS has collected, so every value can be judged against what is typical for that device at that time.

baselines.soldierx.internal • core-sw-01
core-sw-01 · Uplink utilizationTypical rangeCurrent
Mon 00:00Mon 12:00Tue 00:00Tue 12:00Now
Current
62%
Typical for Tue 10:00
48 – 71%
Median
57%
Suggested threshold
84%

Per Device, Interface and Time of Week

Baselines are calculated per device, per interface and per time of week, so a busy Monday morning is not compared with a quiet Sunday night.

Statistics That Resist Spikes

Baselines use measures such as mean, median, minimum, maximum, standard deviation and percentiles. Median and percentile measures are not distorted by short spikes.

Typical Range Beside Every Value

See the typical range for a metric next to its current value and understand at a glance how a device compares with its usual behavior.

Suggested Thresholds

Get suggested alert thresholds for individual devices in place of one fixed number for the whole network, and reduce alerts that fire on normal activity.

Coverage

Metrics With Baselines

CPU utilization
Memory utilization
Temperature
Interface utilization
Interface errors
Interface discards

Explore the AI Engine

From Normal to Unusual

Baselines are the foundation for AI Anomaly Detection. See how unusual behavior is flagged as it starts.