Alerts Practical By Samson Tanimawo, PhD Published Apr 19, 2026 4 min read

Saturation vs Utilization Alerts

Two types of resource alerts. Pick by what they catch.

Utilisation: trailing indicator

CPU at 80%, memory at 70%, disk at 60%. Numbers describing what has been used.

Useful for trends, capacity planning, dashboard panels. Less useful for alerts; by the time utilisation is high, the issue is already present.

Static thresholds rot. The right threshold for one workload is wrong for another. Alerts on utilisation alone produce noise.

Queue depth growing, wait time increasing, throttle events firing. Numbers describing pressure.

Fires earlier than utilisation. CPU at 80% might be fine if queue depth is zero; queue depth growing predicts trouble.

Catches issues before user-visible failure. The monitoring foundation for predictive alerting.

Saturation alerts page first. Queue depth above threshold for 5 minutes.

Utilisation alerts go to dashboards or business-hours notifications. Trends, not pages.

Together they catch different failure modes. Saturation for incipient overload; utilisation for sustained capacity issues.

Database: connection pool wait time (saturation) versus connection count (utilisation).

Network: packet drop rate (saturation) versus bandwidth utilisation (utilisation).

Disk: I/O wait queue depth (saturation) versus disk space used (utilisation).