OTN Mapping — the digital wrapper around every client
A wavelength carries light, but the traffic riding that light is digital — Ethernet, storage, older SONET/SDH, whatever the customer hands over. OTN (Optical Transport Network, ITU-T G.709) is the standard digital wrapper that takes any of those clients, packages them in a fixed structure, adds hardware performance monitoring and error correction, and lets a carrier groom and multiplex them deterministically. This page walks the wrapper layer by layer and explains why operators wrap clients in OTN instead of putting them straight onto a wavelength.
A client is wrapped into an OPU (payload), which is wrapped into an ODU (the switchable path layer that carries monitoring), which is wrapped into an OTU (the line layer that carries framing and FEC), which is what modulates the wavelength. Client → OPU → ODU → OTU → wavelength. Read the rest of the page as unpacking those four boxes.
The OPU / ODU / OTU nesting
OTN wraps in three nested containers, each adding its own overhead for its own job. Think of it as boxes inside boxes: the innermost box holds the customer's bits, and each outer box adds the information the network needs to carry and monitor that box.
OPU — Optical Payload Unit
The adaptation layer. Its job is to fit the client into the container: it carries the client payload plus justification bytes that absorb the small rate difference between the client's clock and the container's clock. The OPU is where "this Ethernet stream" becomes "payload of a standard-rate container."
ODU — Optical Data Unit
The path layer — the box that travels end to end and is the unit the network switches and grooms. It carries the performance monitoring: TTI trail trace, BIP-8 error monitoring, and TCM tandem-connection monitoring. When people say a network "does OTN switching," they mean switching ODUs.
OTU — Optical Transport Unit
The line layer for one optical span. It adds the frame alignment (FAS/MFAS), a general communications channel (GCC0), and — crucially — FEC, the forward error correction that lets the wavelength tolerate noise. The OTU is regenerated hop by hop; the ODU inside it rides through.
The ODU rate ladder
ODUs come in a fixed set of rates matched to common client rates, plus a flexible one. This ladder is worth knowing because grooming and multiplexing decisions are all expressed in these units:
| ODU rate | Nominal capacity | Typical client it carries |
|---|---|---|
| ODU0 | ~1.25G | 1G Ethernet (the base grooming granularity) |
| ODU1 | ~2.5G | legacy 2.5G (e.g. OC-48/STM-16) |
| ODU2 | ~10G | 10G (e.g. OC-192/STM-64, 10GbE WAN) |
| ODU2e | ~10.3G | 10G Ethernet LAN rate (slightly higher than ODU2) |
| ODU3 | ~40G | 40G clients |
| ODU4 | ~100G | 100G clients (100GbE) |
| ODUflex | arbitrary | any rate — sized to the client (e.g. sub-rate or odd-rate services) |
ODUflex is the escape hatch: rather than force an odd-rate client up into the next fixed step and waste capacity, ODUflex is built to an arbitrary size that matches the client, then multiplexed using an integer number of tributary slots. (Capacities above are the standard nominal rates; treat exact bit rates as reference values to confirm against G.709.)
Multiplexing lower ODUs into higher ones — tributary slots
A big win of OTN is sub-lambda grooming: packing several smaller ODUs into one larger ODU that then rides a single wavelength. The larger ODU's payload is divided into tributary slots of fixed granularity (commonly 1.25 G, and 2.5 G in older gear), and each smaller ODU is assigned enough slots to hold it.
GMP vs AMP — two ways to fit the pieces
Mapping a client or a lower ODU into its container has to absorb the small, continuous difference between two clocks. OTN defines two mechanisms:
AMP — Asynchronous Mapping Procedure
The older mechanism. It uses a few fixed justification bytes per frame that can each be "stuffed" or not, absorbing a narrow range of rate difference. AMP works well for the specific fixed client rates it was designed around but is not flexible about arbitrary rates.
GMP — Generic Mapping Procedure
The general mechanism used for the higher rates and for ODUflex. GMP distributes the payload across the container using a calculated count of how many payload words go into each frame, so it can adapt to a wide, arbitrary range of rate differences. GMP is what makes flexible-rate and higher-rate mapping practical.
Beyond 100G — OTUCn and FlexO
The fixed ODU ladder tops out at ODU4 (~100G). For clients and line rates above 100G, G.709 was extended with OTUCn: a scalable line signal built as n × 100G (the "C" is the Roman numeral for 100). An OTUC4, for example, is a ~400G structure; OTUC8 is ~800G. Because such a large signal often will not fit on one optical carrier, FlexO (Flexible OTN) defines how to spread an OTUCn across several manageable interface lanes/carriers and reassemble it.
The overhead — how OTN monitors itself in hardware
A defining feature of OTN is that the monitoring is built into the frame and done in hardware, independent of the client. Every ODU/OTU carries dedicated overhead:
BIP-8 — error monitoring
A Bit-Interleaved Parity check (8 bits) computed over the frame lets each layer count bit errors continuously and precisely, without touching the customer payload. This is how OTN reports link error performance as a hardware fact rather than an inference.
TTI — trail trace
The Trail Trace Identifier is a text tag carried in the overhead naming the expected source/sink of a trail. If the received TTI does not match what a port expects, the network raises a mismatch alarm — this catches miscabling and misconnection quickly.
TCM — tandem connection monitoring
Tandem-Connection Monitoring adds up to six nested monitoring sections inside the end-to-end path. This lets different operators (or different network segments) each monitor just their own portion of a path independently — essential when a circuit crosses multiple carriers and each needs to prove its own segment is clean.
GCC + FEC
GCC (General Communication Channels — GCC0 in the OTU, GCC1/2 in the ODU) are in-band data channels the network uses for its own management/control signaling. FEC, carried in the OTU, is the forward error correction that gives the wavelength its noise tolerance (the coding-gain story from Link Engineering).
Why OTN — versus just putting a client on a wavelength
You could modulate a client straight onto a wavelength. OTN adds a wrapper for four concrete reasons:
1 · Transparent transport of any client
The OPU/ODU structure carries the client's bits and its timing unchanged, so Ethernet, SONET/SDH, Fibre Channel and others all ride the same network transparently. The client comes out the far end exactly as it went in.
2 · Hardware performance monitoring
BIP-8, TTI and TCM give continuous, per-layer, in-hardware error and connectivity monitoring that does not depend on the client protocol — you can prove the transport's health independent of what is riding it.
3 · Deterministic sub-lambda grooming
Tributary-slot multiplexing lets you pack many small services into one wavelength and switch them at known, fixed positions — efficient use of each expensive wavelength, with predictable capacity.
4 · Multi-operator TCM
Tandem-connection monitoring lets each carrier segment on a multi-operator circuit monitor and prove its own portion, which is exactly what is needed to place fault and settle responsibility across a hand-off.
When a service crosses your network and a neighbor's, OTN's TCM is what lets you show "clean through my segment" with hard BIP-8 error counts, and TTI mismatch alarms are your fastest catch for a mis-patched hand-off. Treat the ODU as the thing you provision and monitor; the OTU/FEC and the wavelength are the line underneath it.
The ODU capacities, tributary-slot counts, and n×100G figures here are the standard nominal values used to teach the structure. Exact bit rates, slot arithmetic, and which ODU/OTUCn types a given platform supports must be confirmed against ITU-T G.709 and the Ekinops docs / BOM for your specific hardware.
1. Put the four OTN containers in order from the client outward, and give each one's main job in a few words.
2. A customer hands you a 100G Ethernet service. Which ODU carries it, and roughly how many of them fill a wavelength versus ten 10G services?
3. What is ODUflex for, and which mapping procedure makes it practical?
4. A circuit crosses your network and a partner carrier's. How does OTN let you prove your own segment is clean, and what alarm catches a mis-patched hand-off?
5. Give two reasons an operator wraps a client in OTN rather than modulating it straight onto a wavelength.