A DevOps team polls 400 routers with CLI scraping and SNMP every five minutes, causing high CPU, delayed alarms, and fragile scripts when output changes. They need real-time streaming and vendor-neutral configuration. Which gRPC and gNMI approach best modernizes telemetry and configuration for this estate?
Select an answer to reveal the explanation.
Short Explanation
You can replace fragile screen scraping and slow polling with live, structured updates from your routers. Choose model-driven telemetry over gNMI and gRPC for streaming state and vendor-neutral configuration.
Full Explanation
gNMI over gRPC is the best fit because it combines persistent streaming subscriptions—using sample or on-change modes—with structured Get and Set operations on YANG paths. Streaming model-driven state reduces the CPU burden of polling and the five-minute alarm delay, while common OpenConfig paths replace fragile CLI scraping and support vendor-neutral operation across the 400 routers. Custom gRPC protobuf streaming with vendor-specific extensions fails the vendor-neutral requirement: each platform can require its own schema and tooling. Using gNMI with OpenConfig for configuration but polling operational state retains polling load and delayed detection, so it does not meet the real-time telemetry need. NETCONF over SSH with YANG is a valid model-driven configuration approach and can provide event notifications, but it is not the requested gNMI-over-gRPC approach or its built-in Subscribe modes for streaming telemetry. An exam caveat: gNMI is typically dial-in, with the collector connecting to the router; some gRPC telemetry designs are dial-out, with the router initiating the connection. In practice, test Subscribe with on-change support, common OpenConfig paths, and Set with rollback on routers from two vendors in a lab.