Monitoring DDM in production

Reading DDM once tells you a module's state; trending it tells you its future. Most optical failures announce themselves weeks in advance — as a drifting Rx power, a climbing bias current or a rising pre-FEC BER — to anyone who is recording the numbers.

Bench, CLI or NMS

Where	Tool	Best for
Bench	CodingBox DDM screen, CSV export	incoming inspection, baselining, diagnosing a pulled module
Switch CLI	`show interfaces transceiver` and equivalents (per NOS)	spot checks during troubleshooting
NMS / telemetry	SNMP, streaming telemetry (gNMI), vendor APIs	fleet-wide trending and alerting

Collecting over SNMP

Most platforms expose DDM through the standard ENTITY-SENSOR-MIB (entPhySensorValue, with type/scale/precision per sensor) and through vendor MIBs (Cisco entity sensors, Juniper DOM MIB, Arista, Huawei). Open-source NMS such as LibreNMS auto-discover optical sensors and graph them; Zabbix and Prometheus/Grafana work with the same OIDs or with streaming telemetry. Poll every 1–5 minutes — DDM values change slowly, and modules answer the two-wire bus at a leisurely pace.

The link passport

At installation, record for both ends of every link:

Tx power, Rx power, bias current, temperature (the DDM set);
module vendor / part / serial;
fibre length and, where available, an OTDR trace.

This baseline is what later readings are compared with. Without it, "Rx is −12 dBm" is just a number; with it, "Rx has dropped 2.5 dB since March" is a work order.

Alerting: deltas, not just thresholds

Module thresholds are the maker's limits (Thresholds & alarms); your alerts should be tighter and relative:

Signal	Suggested trigger	Likely meaning
Rx power	drop of 2–3 dB from baseline	dirty/damaged connector, fibre bend, far-end laser fading
Rx power	asymmetric between the two directions	problem on the weak side, not the path
Tx bias	rise of 15–20% from baseline	laser ageing — plan replacement
Temperature	> 65–70 °C or +10 °C from baseline	cooling or density problem
Pre-FEC BER (CMIS/VDM)	climbing towards the FEC limit	first sign on PAM4 links
Module alarm flags	any	backstop — should never be the first alert

Reacting quickly

BFD on important links detects a degraded path in milliseconds, long before routing protocols or users notice.
Track per-lane values on multi-lane modules (Per-lane diagnostics) — a single lane failing is the usual AI/DC failure mode.
When a threshold trips, follow the diagnostic ladder: levels → clean → swap → OTDR.

What DDM cannot tell you

Tx power on cheap modules may be a constant (DDM levels).
DDM sees light, not bits: a link can have perfect levels and still error from dispersion, FEC mismatch or a polarity fault — pair DDM with error counters.
Resolution is 0.1 µW: nothing below −40 dBm is measurable.

In CodingBox

CodingBox is the bench side of this process: a configurable polling interval, a measurement log and CSV export produce the link passport before installation and let a pulled module be compared with its own history in the code database.