Home · Documentation · Technology · Streaming Telemetry

Streaming Telemetry — gNMI, OpenConfig, gRPC

SNMP polling at 60-second intervals is dead. Modern OcNOS deployments stream model-driven telemetry over gRPC with sub-second granularity, OpenConfig-modelled paths, and dial-out collectors — feeding straight into Prometheus, InfluxDB, and Grafana without a translation layer.

Telemetry Pipeline

OcNOS routers stream gNMI Subscribe payloads (gRPC dial-out) to a central collector. The collector demuxes into time-series and search backends, and Grafana renders the dashboards on top.

gNMI streaming telemetry collector pipeline Streaming telemetry pipeline. Multiple OcNOS routers stream gNMI dial-out telemetry over gRPC to a central collector. The collector fans data out to Prometheus, InfluxDB, and Grafana for time-series storage, ad-hoc query, and dashboarding. OcNOS-1 gNMI OcNOS-2 gNMI OcNOS-N gNMI gRPC dial-out Collector Telegraf · gNMIc Prometheus TSDB InfluxDB TSDB Grafana Dashboards gNMI · OPENCONFIG YANG · 1-SEC GRANULARITY · JSON + PROTOBUF

Why streaming telemetry replaced SNMP

SNMP poll cycles miss microbursts, hammer the control plane, and can't deliver a structured, model-driven payload. gNMI Subscribe over gRPC inverts the model — the device pushes typed updates on change or on a sample interval, the collector simply parses the protobuf. OcNOS exposes the full OpenConfig tree plus IETF and IP Infusion native models, and supports both dial-in (collector connects to device) and dial-out (device connects to collector behind NAT) deployments.

The OcNOS streaming telemetry implementation

gNMI Subscribe

ON_CHANGE / SAMPLE / TARGET_DEFINED

Full Subscribe RPC support including ON_CHANGE for state deltas, SAMPLE for periodic, and TARGET_DEFINED to let the device pick the right cadence per leaf.

YANG Models

OpenConfig + IETF + native

OpenConfig interfaces, network-instance, BGP, and platform models; IETF YANG where ratified; native models for OcNOS-specific state. Models are versioned and published.

Dial-out Mode

Behind-NAT routers

gRPC dial-out lets cell-site and edge routers behind NAT push telemetry to a central collector — no inbound port-opening required.

Sensor Coverage

1-second granularity

Per-port counters, queue depths, BGP RIB-in/out, optical DOM, CPU and memory state — all available at 1-second sample intervals or finer where the ASIC supports it.

Encoding

JSON_IETF + Protobuf

Both JSON_IETF and protobuf encodings are supported. Protobuf wins on wire efficiency for high-fanout sensors; JSON wins on debuggability.

Collector Integration

Telegraf + gNMIc

Validated reference pipelines using Telegraf's gNMI input plugin or Nokia's gNMIc, feeding Prometheus, InfluxDB, and Grafana out of the box.

What you get with OcNOS telemetry

  • Standards-aligned paths. OpenConfig leaf paths match what your existing collector already expects — no proprietary YANG translation layer.
  • Sub-second visibility. Microburst detection, ECN-marked counters, and queue depth at intervals SNMP can't reach.
  • Open-source stack ready. Reference dashboards for Grafana, drop-in Telegraf configs, and example Prometheus alerting rules.
  • Single config surface. Same gNMI server speaks Get, Set, and Subscribe — telemetry and config-as-code share one auth and TLS path.
BGP RoCEv2 OcNOS-SP Network Automation

Modernizing your observability stack? Talk to a network architect.

Request a Technical Demo →