Protection & OTN — Surviving a Fiber Cut

Why protection exists

A single dark-fiber path is a single point of failure. For anything that matters — a hospital's imaging link, a bank's trading feed, a data-center interconnect — one backhoe should not take the service down for the hours it takes to find and repair a cut. Protection buys you survivability by pre-building a second path and switching to it automatically. You pay for it in extra fiber, extra ports, and extra cost, so it is a deliberate design choice per service, not a default.

The four schemes at a glance

1+1 protection

The signal is bridged onto two paths at the same time — transmitted down both the working and protect path simultaneously. The receiver selects the better of the two. If the working path fails, the receiver is already getting a good copy on the protect path and switches to it in milliseconds. Nothing has to be signaled across the network first, which is why 1+1 is fast and simple — and why it always consumes the protect path (it's never reusable for other traffic).

1:1 protection

Traffic normally rides only the working path. On failure, both ends switch to the protect path. Because the protect path is idle in normal operation, it is reusable — it can carry lower-priority "extra traffic" that gets pre-empted when a protection switch is needed. You trade a little speed (you must detect the fault and coordinate the switch) for better capacity efficiency. The generalization to shared capacity is 1:N — one protect path backing up N working paths.

SNCP — per-service protection inside OTN

SNCP — subnetwork connection protection — is the OTN world's version of the 1+1 idea applied per service in the digital layer. An individual service (one ODU) is bridged across two diverse paths through the network, and the far end selects the healthy one. The value over raw optical 1+1 is granularity and intelligence: you protect specific services rather than a whole fiber, and the switch decision is made on the quality of that specific digital signal (BIP errors, loss of frame, backward defects) — not just "is there light." Recognize it as "per-service, dual-fed, receiver-selects, inside OTN."

Ring protection

Nodes are wired in a ring. Traffic normally travels one way around; if a span fails, traffic is wrapped (folded back at the nodes adjacent to the cut) or steered (rerouted the long way) to reach the same destination. Rings are efficient because every span does double duty — working path for some services, protect path for others — so you get resilience for many services without a dedicated second path each. Target restoration is the classic ~50 ms.

The ~50 ms switch target and the timers

The traditional carrier target for a protection switch is ~50 milliseconds — fast enough that voice, video, and most sessions ride through without dropping. Two families of timers shape the behaviour around that switch:

Optical-layer vs OTN-layer protection

Protection can live at two different layers, and good designs are explicit about which (and avoid having both fight over the same fault).

OTN — the digital wrapper (ITU-T G.709)

OTN — Optical Transport Network — is a digital wrapper standardized to carry client signals (Ethernet, Fibre Channel, older SONET/SDH) across the optical network in a uniform, manageable way. Two things make it worth understanding: it adds management the client lacks (monitoring + FEC), and it makes signals multiplexable (groom many small services into one big pipe, each independently managed).

The nesting: OPU inside ODU inside OTU

• OPU (Optical Payload Unit) — the client's payload plus the mapping/justification overhead that fits an arbitrary client rate into the container.
• ODU (Optical Data Unit) — the container that carries the payload plus end-to-end path monitoring. This is the "service" object you groom, switch, and protect (e.g. ODU2 for 10G, ODU4 for 100G).
• OTU (Optical Transport Unit) — the ODU plus FEC and section overhead, ready to modulate onto the wavelength for one hop across the line. This is the "on the fiber" framing.

Approximate line rates

Keep these as ballpark figures for reading designs — the exact rates carry OTN/FEC overhead above the client rate, which is why OTU4 (~111.8 G) is more than the 100 GbE it carries.

Rates are approximate/nominal and exclude vendor-specific variants; treat them as "which bucket am I in," not exact figures to quote.

ODUflex, tributary slots, and GMP

FEC — forward error correction and coding gain

FEC is why an optical span can run with a "bad-looking" pre-FEC error rate and still deliver zero errors to the customer: the wrapper carries enough redundancy to correct errors before hand-off. More correction = more net coding gain = more dB of margin, which converts directly into reach.

Add/drop, express, and pass-through services

At any node with a ROADM or add/drop mux, a service can be handled three ways. Knowing which applies to each wavelength at each site is core to reading a network diagram.

Protection switching in action

Put it together. Normal state runs the working path. A span fails. The receiver (already fed on both paths, in 1+1) detects loss of signal or excessive errors and selects the protect path. The customer sees a sub-second hit at most; many services see nothing.

Maintenance without customer impact

Protection is not only for failures — it is the mechanism for hitless maintenance. To work on the working path safely: verify the protect path is healthy, manually switch traffic to protect (a "forced switch" or "manual switch" in the NMS), confirm the service is stable on protect, then do your work on the now-idle path. Reverse when done. This is the optical equivalent of vMotioning VMs off a host before you patch it.