What IP over DWDM means
IP over DWDM puts the coherent wavelength straight into a router port, so the transport layer between two routers is no longer a separate shelf of transponders and muxponders. The fibre, the amplifiers and the mux or ROADM units stay; what disappears is the box that sat in front of them.
This guide covers what the change removes, what has to be kept, the optics used and the line items that decide whether the collapse pays. The modules themselves are covered on the coherent optics page and in the 400G ZR DCI selection guide.
What collapsing the transport layer actually removes
In a classic build, each wavelength needs a transponder card with its own processor and a pair of short grey optics for the client signal from the router. Collapsing the layer deletes that card, its grey optics, its rack unit, its power and its management entry, and the router port drives the wavelength directly.
What stays is the line side: mux and demux units or a ROADM, amplifiers, the supervisory channel and the fibre. The two ends also have to agree the mode and error correction profile, because with no transport card in the middle there is nothing to translate between them.
Why operators collapse the layer
Four gains are usually quoted: fewer boxes, so less power, floor space and spares; fewer ports, since the client port, the grey optic and the transponder port all disappear per wavelength; a shorter path, with a pair of short optics and a card out of the latency chain; and faster turn-up, by inserting and configuring one module.
The cost is on the assurance side. The per wavelength overhead, the performance monitoring and the loopback points that lived in the transport layer are now read from the module and the router, and optical layer protection falls to the routers instead.
Recommended optics for an IP over DWDM link
| Module | Rate and mode | Media / wavelength | Reach class | Where it usually lands |
|---|
| 400G QSFP-DD 400ZR | 400G single carrier | SMF, DWDM C-band tunable | Metro DCI class, around 120 km | Router to router on a short point-to-point span |
| 400G QSFP-DD OpenZR+ | 400G, or lower rate modes | SMF, DWDM C-band tunable | Metro to regional class | Links reaching past the metro ring |
| 400G QSFP-DD or OSFP ZR+ | 400G | SMF, DWDM C-band tunable | Amplified spans of a few hundred km | Regional runs on an existing line system |
| 800G QSFP-DD or OSFP ZR and ZR+ | 800G, or 2 x 400G | SMF, DWDM C-band tunable | Metro DCI class upward to regional | Highest capacity per router port |
| 100G and 200G CFP2-DCO | 100G to 200G | SMF, DWDM C-band tunable | Metro to long haul | Router cards and shelves built around CFP2 |
| Fixed wavelength DWDM parts, 10G and 100G | 10G or 100G per channel | SMF, DWDM grid, fixed channel | Span dependent | Lower rate channels on an external mux |
| Grey LR4 and ER4 parts | 10G to 400G | SMF, 1310 nm or LAN-WDM | 2 to 40 km class | Links with spare fibre and no wavelength needed |
Reach classes are the standard ones for each mode; the working limit is the measured loss on the route. Rate families are listed on the 400G QSFP-DD and OSFP and 800G pages.
The line system, the grid and the alien wavelength question
The router port has to fit the channel plan of the line it feeds. A tunable module can be set to the channel the plan leaves free, and the grid has to match what the mux and ROADM expect. Where a wavelength comes from outside the transport vendor's system, the line owner has to confirm the amplifiers and the channel plan will carry it, tested on one channel first.
Where wavelengths share a pair, the passive side is as in any DWDM build: mux and demux units or a ROADM, amplifiers and a supervisory channel. Our passive OADM and mux modules, 80 channel mux and demux module and 40 channel expansion with amplifiers cover that side of the build.
Fault isolation and what the removed layer used to do
With no transport card in the path the wavelength is monitored through the module and the router, so module telemetry has to be enabled and read: power, temperature and laser bias. Keep the mode and error correction profile identical at both ends and record them per link.
Optical layer protection also gives way to router level reconvergence, so the convergence timers and failure scenarios belong in the build plan, not the handover pack.
Port economics, power and when to keep the transport layer
Count ports and cages first: a collapsed link frees the client and transponder ports and the two grey optics per wavelength, but it moves the module power into the router cage, and a rack at its thermal limit gains less than the inventory suggests. Sparing changes too: spares are held per rate and per mode.
Some links should keep the transport layer. Long haul with regeneration, wavelengths that carry services for several parties, line systems with several equipment vendors and networks with assurance requirements all lose more than they save. Mixed networks are normal: collapsed wavelengths where routers are close, transport cards where the path must be managed as a service.
How to choose, in order
- Decide per route whether the transport card stays or goes.
- Confirm the router port supports the rate, mode and power envelope.
- Check the line system grid, channel plan and launch power.
- Measure the span and choose the mode from the measurement, not from the label.
- Agree the mode and error correction profile at both ends per link.
- Confirm how the wavelength is monitored and who acts first on a fault.
Frequently asked questions
Does IP over DWDM remove the need for a mux?
No. It removes the transponder. The mux or ROADM, the amplifiers and the supervisory channel still carry the wavelengths, unless the link is one point-to-point span.
Can a router port drive a DWDM wavelength directly?
Yes, when the port supports the coherent rate and mode and the thermal envelope allows the module. That is the point of the collapsed shape.
How do we monitor a wavelength with no transport layer?
Through the module and the router. Enable module telemetry, read power, temperature and laser bias, and agree at both sites who checks first.
Is IP over DWDM always cheaper?
No. It saves a card, grey optics, power and rack space per wavelength, but it adds coherent module cost, moves heat into the router and weakens per wavelength assurance. Long haul often keeps the transport layer.
What should be in the request for a collapsed link?
The route and measured loss, the line system and channel plan, the rate and mode, the router platform and firmware, the rack power and airflow limits, and the temperature at each end.
Send the route and measured loss, the line system, the rate and mode and the router platform at each end, and we confirm the module, the grid position and the coding for every wavelength. Related reading: the 400G transponder and muxponder cards and 800G DCO transponder card for the transport layer that stays, and all optics selection guides.
