Telemetry performance comparison between gNMI and SNMP in IP/MPLS network architecture

Abstract

The exponential growth of data traffic in Service Provider environments demands highly accurate and real-time network observability. Traditionally, the Simple Network Management Protocol (SNMP) has been the standard for network monitoring; however, its pull-based polling architecture presents limitations in detecting transient anomalies. This study presents a comparative performance analysis between SNMP and the modern, push and Remote Procedure Call (RPC) based gRPC Network Management Interface (gNMI). By using Infrastructure as Code (IaC) principles, a multi-vendor routing environment comprising Juniper and Arista nodes was virtualized using the Containerlab orchestration framework. The methodology subjected the observability pipeline to deterministic control-plane and data-plane anomalies, including volumetric Distributed Denial of Service (DDoS) bursts, Border Gateway Protocol (BGP) route flapping, and Label Distribution Protocol (LDP) intermittency. The research concludes that gNMI provides high-fidelity visibility (better spikes view) and lower operational overhead (39%-50% less CPU usage) for modern autonomous networks, recommending a phased migration strategy for production environments.