Open Source NMS: Great for Servers and Routers, Not for PON Networks
“Right tool, right job. Observium for switch ports? Sure. PON optics? Forget it.”
We love open source. Our racks hum thanks to Prometheus, InfluxDB, Grafana, and a scatter of Bash that should probably be in Git. Observium spotted its first CRC error at Extreme Labs before some of our engineers had their first coffee. LibreNMS once saved us a Saturday by barking about a flapping uplink.
But when the conversation shifts to monitoring GPON/XGSPON OLTs and millions of ONUs, open-source tools often just give up when they hear the word 'PON'. Here’s why-plus how we still put them to work where they shine.
1. Great Generalists, Bad Fiber Specialists
Open‑source tools speak SNMP, ping, and syslog fluently. PON, not so much.
NeedResult in LibreNMS / ObserviumPer‑ONT optical power graphRoll your own script and hope OIDs stay put.Vendor‑specific LOS/LOF alarmsPossible—after days in the MIB jungle, or parsing multi-line, vendor-specific syslog formats for LOS, LOF, or Dying Gasp events.Split‑ratio heat mapOut of scope.Aggregated OLT capacity across six vendorsPray and grep.
The problem isn’t philosophy; it’s schema. PON is a jungle of proprietary tables and TLVs. Teaching each open‑source project every vendor’s idea of “good power” is a full‑time job.
2. The DIY Tax (a.k.a. Sysadmin Hours)
Open source is “free” the same way a puppy is free:
- Install day → Get your favorite Linux distro, install a dozen dependencies, compile from source, and spend a day wrestling with config files and obscure acronyms.
- Month 3 → RRD files balloon to hundreds of gigabytes, graphs crawl, MySQL chokes, and you start googling partitioning tricks.
- Year 1 → Upgrade night breaks half the plugins; you debug until dawn.
If your NMS collapses when fiber cuts, nobody cares that the licence was $0.
3. Scale Is a Four‑Letter Word
PON math is brutal: 4,000 OLTs × 8 PON ports × 50 ONUs ≈ 1.6 million data points. RRDtool chokes. SNMPwalks hit rate‑limits. Grafana struggles to keep up with high-frequency intervals and large-scale metrics from PON networks, often causing slowdowns and graphing glitches.
Horizontal sharding? LibreNMS has distributed polling, yes, but try coordinating poller drift across five data centres. One misfired cron and your SLA reports look like Swiss cheese.
4. Where We Still Say “Yes” to Open Source
- Servers & VMs → Prometheus + Grafana. Perfect.
- Edge switches & BGP sessions → LibreNMS, because its routing modules rock.
- Ad‑hoc analytics → InfluxDB + Python notebooks. Cheap and cheerful.
Open source is unbeatable when the domain is well‑documented, the pace of change is sane, and vendors converge on common protocols. PON is still the Wild West.
5. One Dashboard, Zero Silos
We’ve ditched the patchwork. NetSense now ingests telemetry from everything with a MAC address-OLT, ONU, switch, router, server, BNG/bRAS, etc. A unified data platform with scalable storage, one alerting pipeline, one place to correlate a sudden drop in ONT count on a PON port with a high number of Dying Gasp reasons-indicating not a network fault but a localised power outage, removing false alarms and focusing on what matters.
- Unified inventory – Switches, routers, and PON ports all live in the same tree-one unified network inventory without the mess.
- Single alert stream – Our own Alerting module (Alerta) or flexible options like Webhook or Telegram ensure a single alert stream without duplicated rules or messy overlaps.
- Cross‑domain analytics – Spot a rising rack temperature before it drags RX power into the red.
Bottom line: if it speaks SNMP, gRPC, syslog, streaming telemetry—or even plain CSV—NetSense collects it, stores it, builds historical reports, correlates events, and provides insights. No jump‑tabs, no context‑switches.
Conclusion: Use the Right Tool for the Right Job
Observium, LibreNMS, Prometheus-they’re hammers. PON is a screw. You can pound, or you can grab a screwdriver.
Need a screwdriver? Talk to us. We’ll show you how NetSense was designed for fiber networks (FTTH, FTTx, etc) from day one-and how it plays nicely with the open‑source kit you already trust.
