Summary

  • MPLS-TP linear protection is a PSC state-control function inside an existing protection domain. It coordinates selectors; it does not create paths, reserve capacity, authenticate messages by itself, or give either endpoint unconstrained routing authority.
  • Five local trigger sources feed the decision: operator commands, server-layer indications, control-plane indications, OAM indications and expiry of Wait-to-Restore (WTR). A far-end PSC request is also visible, but its effect remains bounded by the configured architecture and state rules.
  • Protection spends capacity assigned in advance for speed. Restoration may reroute over available capacity, is slower and can fail when resources are unavailable.

How the decision is made

The local-request logic selects the highest-priority current local request. The control logic then cross-checks that request against the far-end request, determines the endpoint selector action, establishes the protection-domain state and generates the outgoing PSC message. The message is coordination input, not a general routing command. Recognizing a Channel Type does not, by itself, authenticate or authorize the sender.

A practical priority table is:

Priority, high to low Request or condition Operational meaning
1 Lockout of protection Keep protection unavailable; operator intent dominates switching.
2 Forced Switch Move traffic as commanded despite ordinary fault logic, subject to the mode.
3 Signal Fail Protect against a qualifying failure condition.
4 Signal Degrade In an enabled operational mode, protect against degraded service.
5 Manual Switch Move traffic for planned work or controlled intervention.
6 WTR / no request Delay or permit return according to reversion rules.

The exact effective result still depends on the state machine and provisioning. The states express conditions such as normal traffic on working, traffic on protection, a requested or active lockout, forced or manual switching, and WTR before reversion. A message generator advertises the locally selected request and relevant state so the far end can coordinate; it does not expand the domain.

The effects differ by architecture. In 1:1 protection, working and protection resources are associated and the selectors coordinate which path carries traffic. In bidirectional 1+1, both ends switch together so the same working or protection path is selected in both directions. In unidirectional 1+1, each direction can select independently; crucially, a remote request must not trigger a switch in that unidirectional arrangement. That simplicity can require more dedicated capacity and does not remove the need for consistent provisioning.

Revertive operation returns traffic to working after the condition clears, but WTR deliberately delays that return so an intermittent fault can stabilize. Non-revertive operation can leave traffic on protection until another qualifying request changes the state. Lockout, Forced Switch and Manual Switch are operator controls with different priorities and purposes; Signal Fail is a detected condition, not a diagnosis of root cause or proof of an SLA breach.

RFC 7271 adds capability advertisement and distinguishes PSC and APS modes. Advertisement verifies compatible capability sets; it is not negotiation and does not prove that both endpoints were provisioned correctly. Its operator-oriented behavior includes Freeze, modified priorities, changes to non-revertive behavior, MS-W, signal-degrade protection and exercise commands. Those controls make the operating contract more explicit, while also adding choices that must be tested at both ends.

Protection and restoration are not interchangeable. Protection reserves route and resources before failure, enabling coordinated switching but imposing capacity and consistency costs. Restoration can seek available capacity and reroute after failure; it is slower and is not guaranteed if resources are unavailable. A counterfactual network relying only on restoration trades pre-assigned speed for resource uncertainty. A network allowing uncoordinated local switching risks endpoint disagreement. A simple unidirectional 1+1 choice may reduce coordination dependence, but it spends dedicated resources and has the stated remote-request limitation.

Sources