InfluxData has released an open source, runnable reference architecture for network telemetry built on a 5-node InfluxDB 3 Enterprise cluster. The project simulates an 8×16 Clos data-center fabric ingesting ~10,000 points/second and demonstrates several advanced patterns: role-separated multi-node deployment (ingest, query, compact, and process nodes), Python Processing Engine plugins with cross-node write-back, Last Value Cache and Distinct Value Cache for sub-millisecond lookups, and per-table retention policies. Three browser-to-database integration patterns are shown side by side with live latency badges: server-side SQL via FastAPI, browser-direct SQL using DVC table-valued functions, and request-triggered Python plugins. The entire stack runs locally via docker compose and includes scenario injectors for congestion hotspot and east-west burst events. Production wiring guidance covers Telegraf input plugins for SNMP, gNMI, NetFlow/IPFIX, and vendor REST APIs.
Table of contents
What is the network telemetry reference architecture?What’s in the stackThe features it’s actually showing youThree integration patterns, side by sideWhere to wire in real network dataTry it yourselfA note for AI agentsResourcesSort: