Tutorial 8: Network Telemetry, YANG, gNMI, and Model‑Driven Management

COMP347 Unit 8 – Network Management and Network Operations

Table of Contents

• Learning Objectives
• Overview
• Technical and Theoretical Content
  • 1. Introduction to Network Telemetry
  • 2. YANG Data Modeling Language
  • 3. gNMI: gRPC Network Management Interface
  • 4. Model‑Driven Telemetry and Streaming
  • 5. Telemetry vs. SNMP: A Comparative Analysis
  • 6. Telemetry Data Pipelines and Processing
  • 7. OpenConfig and Native YANG Models
  • 8. Security and Access Control in gNMI
  • 9. Operational Considerations and Best Practices
  • 10. Case Studies
• Quiz (42 questions)
• Exercises (20 problems)
• Homework (16 assignments)
• Summary

Learning Objectives

After completing this extended tutorial, you should be able to:

• Define network telemetry and articulate its advantages over traditional polling.
• Describe the YANG data modeling language, its structure, and common data types.
• Explain the gNMI protocol, its operations (Capabilities, Get, Set, Subscribe), and message formats.
• Differentiate between streaming telemetry and periodic polling, and understand their trade‑offs.
• Compare and contrast SNMP (with SMI/MIB) and telemetry (with YANG/gNMI) in terms of architecture, data models, and use cases.
• Design a telemetry data pipeline for collecting, storing, and analysing streaming network data.
• Identify the role of OpenConfig models and their importance for vendor‑neutral management.
• Evaluate security considerations for telemetry systems, including TLS, authentication, and authorization.
• Analyze real‑world deployments of telemetry in cloud and service‑provider environments.
• Apply model‑driven management concepts to automate network operations.

Overview

Network telemetry is the practice of continuously collecting high‑frequency, structured data from network devices, enabling real‑time monitoring, analytics, and automation. Unlike the traditional SNMP polling model (pull‑based, low‑frequency, and limited data), modern telemetry uses a push model with rich data models (YANG) and efficient protocols like gRPC and gNMI. This tutorial provides a comprehensive, in‑depth exploration of network telemetry, YANG, and model‑driven management.

We begin by defining telemetry and its key drivers: the need for higher frequency, richer data, and scalability. We then delve into the YANG data modeling language, its syntax, and how it represents configuration and operational state. The gNMI (gRPC Network Management Interface) protocol is examined in detail, covering its operations (Capabilities, Get, Set, Subscribe) and its use of gRPC for efficient, bi‑directional streaming. We contrast telemetry with SNMP, highlighting the architectural and functional differences. The tutorial also covers telemetry data pipelines, the OpenConfig initiative, security considerations, and practical deployment strategies. Case studies illustrate telemetry in large‑scale networks.

Technical and Theoretical Content

1. Introduction to Network Telemetry

Network telemetry refers to the automated, high‑frequency collection of data from network devices to enable real‑time monitoring, analytics, and automation. It is driven by the need for:

• Higher frequency: Sub‑second data collection for near‑real‑time decision making.
• Richer data: More than just interface counters; includes QoS, routing, and application‑level metrics.
• Scalability: Efficient support for thousands of devices and millions of data points.
• Automation: Enabling closed‑loop control and intent‑based networking.

Telemetry is typically push‑based: devices stream data to a collector without being polled, reducing overhead and enabling higher frequencies.

2. YANG Data Modeling Language

YANG (Yet Another Next Generation) is a data modeling language used to model configuration and state data for network devices. It was defined in RFC 6020 (updated by RFC 7950). YANG models are hierarchical and can represent:

• Configuration: Writable data that defines how the device should behave.
• State: Read‑only operational data (counters, status).
• RPCs: Remote procedure calls.
• Notifications: Event‑driven messages.

YANG syntax uses containers, lists, leafs, and leaf‑lists. It supports data types like integer, string, enumeration, and more complex ones like union and leafref.

Example YANG snippet:

container interfaces {
    list interface {
        key "name";
        leaf name { type string; }
        leaf enabled { type boolean; }
        leaf ip-address { type inet:ipv4-address; }
        container state {
            config false;
            leaf oper-status { type enumeration { enum up; enum down; } }
            leaf in-octets { type yang:counter64; }
        }
    }
}
        

YANG modules are used by protocols like NETCONF and RESTCONF, and also form the basis for gNMI telemetry.

3. gNMI: gRPC Network Management Interface

gNMI (gRPC Network Management Interface) is a modern, protocol‑buffers‑based management protocol defined by the OpenConfig working group. It uses gRPC (high‑performance RPC framework) and provides:

• Capabilities: Query the device’s supported models and features.
• Get: Retrieve current values of configuration and state data.
• Set: Modify configuration.
• Subscribe: Stream telemetry data (push) with multiple subscription modes: ONCE (sample now), POLL (on‑demand), and STREAM (continuous).

gNMI messages are encoded in Protocol Buffers (protobuf) over HTTP/2, providing high efficiency, multiplexing, and flow control.

Subscription types:

• STREAM: Periodic or on‑change updates.
• POLL: Client sends a request, device responds once.
• ONCE: One‑time sampling.

Paths in gNMI are defined using XPath‑like notation (e.g., /interfaces/interface[name='eth0']/state/oper-status).

4. Model‑Driven Telemetry and Streaming

Model‑driven telemetry leverages YANG models to define the data to be streamed. Devices export data based on the model, and collectors interpret the data using the same model. This ensures consistency and reduces integration effort.

Streaming telemetry typically involves:

• Subscriptions: Devices are configured with subscription policies (paths, frequency).
• Encoding: Data is encoded in GPB (Google Protocol Buffers) or JSON.
• Transport: gRPC or sometimes UDP (for lighter telemetry).

Advantages: lower latency, higher throughput, and richer data compared to SNMP polling.

5. Telemetry vs. SNMP: A Comparative Analysis

While SNMP remains widely deployed, telemetry is increasingly favored for modern, large‑scale, and automated networks.

6. Telemetry Data Pipelines and Processing

A telemetry pipeline consists of:

• Data sources: Network devices streaming data via gNMI.
• Collectors: gNMI clients that receive and decode streams.
• Message bus: Kafka or similar for decoupling and scalability.
• Stream processors: Flink, Spark, or custom processors for real‑time analytics and anomaly detection.
• Storage: Time‑series databases (InfluxDB, Prometheus) or data lakes.
• Visualization: Grafana, Kibana, or custom dashboards.

Considerations: data volume, latency, retention, and the need for downsampling.

7. OpenConfig and Native YANG Models

OpenConfig is a vendor‑neutral initiative that provides standard YANG models for common network functions (interfaces, BGP, routing, etc.). Using OpenConfig models enables multi‑vendor interoperability and reduces operational complexity. Many devices support both OpenConfig and vendor‑native models.

8. Security and Access Control in gNMI

gNMI uses TLS for encryption. Authentication can be through certificates (mTLS) or username/password over TLS. Authorization is typically handled by the device based on the authenticated user and role‑based access control (RBAC) to specific YANG paths. gNMI also supports secure channels for sensitive configuration operations.

9. Operational Considerations and Best Practices

• Subscription strategy: Balance frequency and data volume; use on‑change for static data, periodic for counters.
• Device capacity: Ensure devices can handle the telemetry load (CPU, memory).
• Network bandwidth: Avoid saturating management links.
• Collector scalability: Use horizontal scaling and load balancing.
• Data modeling: Prefer OpenConfig models for interoperability.
• Testing: Validate telemetry pipelines in a staging environment.

10. Case Studies

• Cloud provider: Uses gNMI streaming to collect interface counters and BGP state from thousands of routers, feeding into a real‑time monitoring system that detects traffic anomalies and triggers auto‑scaling.
• Service provider: Migrates from SNMP polling to telemetry for network performance monitoring, reducing polling load by 90% and achieving sub‑second visibility.
• Enterprise: Deploys OpenConfig models for multi‑vendor network management, using gNMI to both monitor and configure devices from different vendors through a unified platform.

Quiz (42 Questions)

All answers are hidden; click Show Answer to reveal.

Exercises (20 Applied Problems)

Sample solutions are hidden – click to reveal.

container interface {
    leaf name { type string; }
    leaf enabled { type boolean; }
    leaf speed { type uint64; }
    leaf oper-status {
        type enumeration { enum up; enum down; }
    }
}
            

{
    "update": [
        {
            "path": "/system/config/hostname",
            "val": "router-core-01"
        }
    ]
}
            

Homework (16 In‑Depth Assignments)

Sample answers are hidden; use them to guide your study.

Summary

This extended tutorial has provided a comprehensive exploration of network telemetry, YANG, gNMI, and model‑driven management. We covered the foundational concepts of telemetry, the YANG data modeling language, and the gNMI protocol with its operations and subscription models. A detailed comparison with SNMP highlighted the advantages of telemetry in terms of frequency, data richness, and scalability. We also examined telemetry data pipelines, OpenConfig models, security considerations, and operational best practices.

The quiz, exercises, and homework assignments are designed to reinforce understanding and develop practical skills for deploying and managing telemetry systems. As networks evolve towards automation and intent‑based operations, telemetry and YANG play a pivotal role in providing the high‑quality data needed for closed‑loop control and predictive analytics. The next tutorial will explore Configuration Management and Change Control, focusing on maintaining consistent device configurations.