Telemetry-driven and closed-loop automation
Requirement / business driver
Move from open-loop "push config" to closed-loop: the network observes itself and acts - the bridge from automation to assurance.
The loop
Streaming telemetry (Network management - traditional vs model-driven) feeds analytics that detect a condition, decide a response, and act (push a change) - then verify. Variants:
- Event-driven automation - a telemetry event/threshold triggers a remediation workflow.
- Intent-based networking (IBN) - a controller translates business intent -> device config -> continuous assurance, flagging or correcting deviations from intent.
- AIOps - ML over telemetry/logs for anomaly detection and (eventually) auto-remediation (see AI-assisted operations - AIOps in the elective).
Recommendation / justification
Design the observability + automation loop together: rich streaming telemetry as the sensor, a policy/intent engine as the brain, and gated automation as the actuator (with human approval where blast radius warrants). Start with *closed-loop detection/alerting and graduate to auto-remediation* only where it's safe and tested. Justify by operational scale and the speed/consistency of response - while respecting blast radius.
What would change this (mid-scenario twist)
- Large/complex estate, slow manual response -> closed-loop detection + selective auto-remediation.
- High-risk actions -> keep a human in the loop (assisted, not fully autonomous).
Validation checks
- Is there a sensor -> decide -> act -> verify loop, not just dashboards?
- Are auto-remediations gated by risk/blast-radius (human approval where needed)?
IPv6 / dual-stack note
Telemetry and assurance must cover both families; intent/assurance that only watches v4 misses v6 faults.
Related
- Network management - traditional vs model-driven
- CI-CD pipelines for networks (NetDevOps)
- AI-assisted operations - AIOps
Spaced repetition
Describe a closed-loop automation cycle.
Streaming telemetry (sense) -> analytics detect a condition -> decide a response -> act (push change) -> verify.
Intent-based networking translates [...] (flagging/correcting drift from intent).
Intent-based networking translates business intent into device config and provides continuous assurance (flagging/correcting drift from intent).
The safe adoption path for closed-loop is [...].
The safe adoption path for closed-loop is start with detection/alerting, then add auto-remediation only where tested and low-risk (gate high-blast-radius actions with human approval).
Sources
- Cisco Press, Network Programmability and Automation; Cisco DNA/Catalyst Center (IBN) and AIOps references.