SNMP polling is reaching its limits. Streaming telemetry with gNMI delivers real-time, event-driven network observability. Here's how to migrate without losing visibility.
SNMP has served us well for 30 years, but its pull-based polling model introduces inherent delays. At 5-minute intervals, you might miss a 30-second microloop or a brief traffic spike that caused packet loss. Streaming telemetry pushes data from the device to your collector in near-real-time — typically at 10-second or even 1-second granularity.
Why Migrate?
- Resolution: 1–10 second intervals vs. 5–15 minute SNMP polls
- Efficiency: gNMI/gRPC uses binary protobuf encoding — 10× more efficient than SNMP's ASN.1/BER
- Structured data: YANG models provide consistent, typed, hierarchical data — no more OID hunting
- Subscriptions: Dial-in or dial-out models; event-driven notifications for state changes
Migration Steps
- Identify your top 20 SNMP-polled metrics (interface counters, CPU, memory, BGP state)
- Map each to the corresponding YANG path (OpenConfig or vendor-native)
- Deploy a telemetry collector (Telegraf + InfluxDB, or gNMIc + Prometheus)
- Run SNMP and telemetry in parallel for 30 days to validate parity
- Retire SNMP polling for migrated metrics; keep SNMP traps for legacy devices
Start with interface statistics — it's the highest-value, lowest-risk migration target. Once your dashboards are fed by streaming telemetry, you'll never go back to SNMP polling.