NetOp ships settings tuned to cover everything without flooding you. That is a deliberate starting point, not a finished configuration. This article covers the six levers available for tuning, and how to tell whether tuning is working.
Before you tune: settings are per site
Anomaly thresholds, incident categories, and escalation timers apply to every site under a customer. Quietening one noisy site by loosening a threshold loosens it everywhere that customer operates.
This cuts both ways. A retail chain with 200 thin sites and a single-site professional services firm should not share a configuration, even though the defaults would give them one. Review per customer, never globally.
The six levers
1. Threshold level. Move the line to match reality for that customer rather than the factory default.
2. Duration in the hour. The metric must accumulate time beyond the line within the rolling hour before anything is raised. This is what already absorbs brief, self-correcting spikes — often the answer when someone asks why a short spike produced nothing.
3. Multiple severities. Two lines on one metric — an early band and a hard ceiling — turn a single loud alert into a warning followed by a verdict.
4. Escalation timers. Let operational anomalies sit at Medium until they prove persistent. Configured per device class, slower for VPN and faster for firewalls and switches.
5. Port roles. Mark user-facing ports so interface-down is not reported, and mark ISP and interconnect ports so they are continuously monitored. This is the single biggest noise reduction available on switch-heavy sites.
6. Notification topics. Subscribe each recipient only to the categories they act on, and filter by severity. Metric notifications are off by default — turn them on deliberately.
The trade you are making
Every step towards aggressive produces more anomalies and opens more incident tickets. That is sometimes exactly right — a customer with a hard SLA may want it — and sometimes just noise. Make the trade knowingly rather than by drifting.
Configuration that quietly creates noise
Three settings cause more phantom alerts than any threshold:
- Missing bandwidth limits. Throughput is monitored as a percentage of configured bandwidth. Leave the limit blank and the metric cannot be used for monitoring at all — NetOp displays it in statistics only, and no utilisation incident will ever open.
- Untagged SLA profiles. An untagged profile scores traffic against the wrong expectations and reports problems that are not there.
- Wrong device-to-site attribution. A device in the wrong site either joins incidents it has nothing to do with or misses the one it caused. Fix it in Settings › Site Management rather than compensating for it mentally — reassigning takes seconds and mental compensation does not survive a handover.
What good looks like
The measure is anomaly volume falling while real incident numbers hold steady. The Incidents Overtime widget is the one that shows this.
If both fall together, you have tuned out real faults. Stop and audit what you changed.
A tuning routine
Weekly: thirty minutes of tuning owned by a named person. Review new sites and devices for correct site attribution, check the Incidents Overtime trend, and confirm scheduled reports actually generated.
Monthly: review thresholds against what actually happened, review notification recipients (people leave and addresses die), and check Unmonitored Devices for coverage gaps.
Related articles
- How anomaly detection works
- Troubleshooting common symptoms
Comments
0 comments
Please sign in to leave a comment.