Establishing a Baseline
A network baseline captures metrics during normal, healthy operations: bandwidth utilization per link, CPU and memory usage per device, interface error rates, average latency to key destinations, typical packet loss, and common traffic patterns. Baseline collection should span multiple time periods — different times of day, days of week, and business seasons — to capture natural variation.
When to capture baselines: after initial deployment (establishes 'normal' before changes). After significant changes (upgrades, new applications, topology changes). Periodically (quarterly or annually to account for organic growth and change). Baselines become stale over time as the network evolves — regular updates are important.
Using Baselines for Troubleshooting
Deviation from baseline triggers investigation. If interface utilization is normally 30% but spikes to 95% at 2am, that warrants investigation — it could be a backup job, malware, or a rogue device. Baselines provide context: a 10ms ping to the server is fine if baseline is 5ms (slight increase), but concerning if baseline is 1ms (10x increase).
Trending: analyzing baseline data over time reveals capacity issues before they cause problems. If interface utilization grows 5% each month, capacity planning can predict when to upgrade. Anomaly detection: monitoring systems compare real-time metrics against the baseline and alert when significant deviations occur. Modern tools use machine learning to automatically identify anomalous patterns.