Summary

  • RFC 6373 requires a control-plane design for dynamic provisioning and restoration, but using it remains the network operator’s choice. Static provisioning must continue to work without it.
  • MPLS-TP separates control, management and data planes. A healthy out-of-band control channel is not evidence that the associated data channel is healthy.
  • When management and control both provision LSPs, ownership must be explicit and transferable. Management must receive reliable results and retain a view of each control-plane LSP’s operational life.

Automation with a boundary

The framework assigns different mechanisms to different objects. GMPLS RSVP-TE signals MPLS-TP LSPs; targeted LDP signals pseudowires; OSPF-TE or ISIS-TE with GMPLS extensions can distribute traffic-engineering information. These protocols create a dynamic option. They do not abolish static provisioning, and RFC 6373 explicitly says use of the control plane is an operator choice.

That distinction changes the leadership question. The issue is not whether automation is technically available, but which plane currently owns a path, which operation it may perform, and which evidence another plane must receive. A management-created LSP may be handed to control-plane ownership, and a control-plane LSP may be handed back. Without a recorded handover, two systems can reason from incompatible authority states even when each has internally consistent data.

Plane separation creates a second boundary. Control, management and data planes may use independent addresses and different topologies. Control traffic can be in-band, out-of-band on aligned resources, or carried across an independent topology. An independent control channel improves survivability, but its reachability cannot prove the state of the data path. The inverse also matters: a static data path can remain valid when no control plane is present.

RFC 6373 also limits what “OAM control” means. The control plane does not perform OAM; it can instantiate, configure and coordinate OAM functions with the LSP lifetime. Recovery commands can be invoked from management and executed through control-plane mechanisms, commonly at the ingress. That makes the command path auditable only if ownership, trigger, result and affected LSP are correlated.

Concrete control fixture

For every transport path, record an immutable LSP identifier, provisioning mode (static or dynamic), current owner (management or control), last handover receipt, signaling protocol, control-channel topology, data-plane validation timestamp, OAM lifecycle state and recovery-command history. A handover is complete only when the receiving plane acknowledges ownership and the management view records the result.

Test three counterfactuals. First, withdraw the control adjacency while leaving a static LSP forwarding: the path must not be declared invalid merely because signaling is absent. Second, break the data path while keeping an independent control channel healthy: the green adjacency must not be promoted into forwarding evidence. Third, interrupt ownership transfer: neither plane should perform an uncorrelated change until authoritative ownership is restored.

The control plane itself requires survivability, including graceful restart and hot-redundant configurations. This protects the automation substrate; it still does not replace data-plane OAM or management records. Existing MPLS and GMPLS standards should be reused where possible, while functions not needed by a particular MPLS-TP network may remain unused.

The beneficiaries are operators who need dynamic setup and recovery without surrendering static operation or supervision. The cost is explicit state: ownership registers, identifiers, cross-plane notifications, operation logs, reconciliation and resilient controllers. RFC 6373 supplies no evidence about vendor adoption, deployment prevalence, incident reduction, convergence performance, operating cost or commercial outcomes. Those remain unknown; no allegation is established.

Sources