Mist AI and Machine Learning Concepts
This domain tests the Mist AI vocabulary: Service Level Expectations, classifiers, anomaly detection, and how Marvis turns telemetry into a diagnosis. Questions often ask you to distinguish a metric from a root cause.
SLEs are the measurement layer
Wi-Fi Assurance defines Service Level Expectations such as coverage, capacity, roaming, AP health, client health, time to connect, and throughput. Each SLE gets a 0–100 score and a success rate. You set thresholds; Mist measures whether user experience meets them, not whether the AP is merely up.
Classifiers are the diagnosis layer
An SLE failure is broken into classifiers: DHCP failure, DNS failure, auth failure, ARP failure, IP assignment, sticky client, and similar. Classifiers tell you which root cause to chase; the SLE score tells you how bad it is.
Where ML fits
Mist applies machine learning to classify failures from telemetry, detect anomalies against a learned baseline, and correlate symptoms across APs, clients, and switches. Marvis adds natural-language query and generates prioritized Marvis Actions with recommended fixes. Marvis Minis run synthetic client tests for DHCP, DNS, auth, and roaming from APs to catch problems before users report them.
Worked example: the coverage SLE drops from 96 to 71 on 5 GHz on one floor. Only one band and one floor points to RF, not global auth or DHCP - check transmit power, channel width, and new obstructions. The classifier shows clients at −74 dBm, so the fix is RF design, not a RADIUS change.
Traps: treating the SLE score as a configurable setting; confusing classifier names with SLE names; assuming Marvis is only a chatbot; reading correlation as root cause.