Skip to content

Model-Driven APIs in Network Automation

Model-Driven APIs in Network Automation: gNMI, RESTCONF, and the Future of Device Management

This post is part of our ongoing series on network automation best practices, grounded in the PRIME Framework and PRIME Philosophy.

Why This Blog Exists

Legacy CLI and SNMP are giving way to model-driven APIs like gNMI and RESTCONF. This post explains what they are, why they matter, and how to use them for scalable, vendor-neutral automation.

What are Model-Driven APIs?

  • APIs based on YANG data models (standardized, vendor-neutral)
  • gNMI (gRPC Network Management Interface): high-performance, streaming, and config/state access
  • RESTCONF: RESTful API for YANG models, easy to use with HTTP/JSON
  • Enable structured, programmatic device management and automation
  • Replace legacy CLI scraping and SNMP with reliable, scalable interfaces

Why Model-Driven APIs? (Benefits & Use Cases)

  • Consistency: Same data model and API across vendors (multi-vendor automation)
  • Reliability: Fewer parsing errors, less breakage on upgrades
  • Speed: Bulk config/state retrieval, streaming telemetry
  • Intent-based automation: Express desired state, not just CLI commands

Common Use Cases:

  • Automated config deployment and validation
  • Real-time state monitoring and alerting
  • Closed-loop remediation (auto-fix drift)
  • Compliance and audit reporting

YANG Data Models: The Foundation

  • YANG defines the structure of config/state data (interfaces, BGP, QoS, etc.)
  • Vendors and standards bodies publish YANG models (OpenConfig, IETF, Cisco)
  • Tools: pyang, yang-explorer for browsing and validating models

Example 1: Using pygnmi for gNMI (Config/State)

1
2
3
4
5
6
7
8
from pygnmi.client import gNMIclient
with gNMIclient(target=('router', 57400), username='admin', password='pass', insecure=True) as gc:
  # Get interface state
  result = gc.get(path=['/interfaces/interface[name=Ethernet1]'])
  print(result)
  # Set config example
  set_result = gc.set(update=[('/interfaces/interface[name=Ethernet1]/config/enabled', False)])
  print(set_result)

Example 2: Using RESTCONF with Python (requests)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
import requests
import json

url = 'https://router/restconf/data/ietf-interfaces:interfaces/interface=Ethernet1'
headers = {
  'Accept': 'application/yang-data+json',
  'Content-Type': 'application/yang-data+json',
}
auth = ('admin', 'pass')
response = requests.get(url, headers=headers, auth=auth, verify=False)
print(json.dumps(response.json(), indent=2))

Advanced Patterns: Abstraction, Error Handling, and Portability

  • Build Python classes/functions to abstract API calls (hide protocol details)
  • Handle errors and API version changes gracefully
  • Use environment variables or vaults for credentials
  • Track vendor support and API changes over time

PRIME in Action: Transparency, Ownership, and Measurability

  • Document API usage, dependencies, and data models
  • Build abstraction layers for portability and future-proofing
  • Track API changes, vendor support, and automation outcomes
  • Use model-driven APIs for compliance, audit, and reporting

Summary: Blog Takeaways

  • Model-driven APIs are the future of scalable, reliable, and vendor-neutral network automation
  • Use gNMI and RESTCONF for structured config, state, and telemetry
  • Build abstractions and track changes for long-term success
  • PRIME principles ensure safe, transparent, and measurable adoption

📣 Want More?