LDP Extension for Inter-Area Label Switched Paths (LSPs)
Extracted elements (18)
Each LSP established using the Longest-Match procedure still strictly follows the shortest path(s) defined by the IGP, preserving traffic engineering consistency with IP routing.
For LER failure detection when IGP aggregates prefixes (so specific LER addresses are no longer advertised), implementations can rely on LDP ordered distribution to propagate unreachability, or advertise LER reachability in IGP control-plane-only (without creating FIB entries).
Standard LDP requires exact match between FEC element and RIB entry, preventing inter-area LSP setup when ABRs aggregate IP prefixes. This document introduces Longest-Match Label Mapping to allow LSPs across IGP areas without leaking specific PE loopback addresses into all areas.
This procedure reduces IGP LSDB size and FIB entries (improving IP router scaling) but does not reduce LFIB entries, so it is insufficient alone to solve MPLS router scaling. Additional MPLS hierarchy in LDP would be needed for that, which is out of scope.
Three existing solutions for inter-area LSPs each have drawbacks: IGP route leaking defeats the purpose of area hierarchy; BGP/MPLS hierarchy requires BGP on all ABRs and prevents IP/LDP Fast Reroute; inter-area RSVP-TE is undesirable when LDP is already deployed and TE is not needed. A pure LDP-based longest-match solution avoids all these drawbacks.
For successful end-to-end inter-area LSPs, this specification must be implemented on all LSRs in all areas where IP aggregation is used. Deployment can be incremental per-area; during upgrade, ABRs should continue advertising specific prefixes until all LSRs in the area are upgraded.
LSRs compliant with this document are backward compatible with RFC 5036 LDP implementations. If a non-compliant LSR exists on the path, the LSP stops at that router with no other adverse effects.
An LSR receiving a Label Mapping message for FEC1 SHOULD use the label for MPLS forwarding if its routing table contains an entry that matches FEC1 (longest match) and the advertising LSR is a next hop to reach FEC1.
If the LSR uses the received label for forwarding, it SHOULD advertise the received FEC element FEC1 (not the aggregated RIB prefix) and a label to its LDP peers.
The Longest-Match Label Mapping Procedure SHOULD be possible to activate/deactivate by configuration, and it SHOULD be deactivated by default. It MAY be possible to activate it on a per-prefix basis.
When a new prefix appears in the RIB, the LSR MUST check if this prefix is a better (more specific) match for some existing FECs, potentially changing the next hop LSR and NHLFE for those FECs.
When a RIB prefix disappears, the LSR MUST check all FECs using that prefix as best match. If another RIB prefix matches, LDP MUST use it; otherwise the LSR MUST remove the FEC binding and send a Label Withdraw message.
When the next hop of a RIB prefix changes, the LSR MUST change the NHLFE of all FEC elements using that prefix.
FECs selected by this procedure use ordered label distribution mode. On LER failure, reachability removal propagates via Label Withdraw messages as defined in LDP section A.1.5.
LDP re-advertises the specific FEC element FEC1 received from downstream, not the aggregated prefix found in the IP RIB during the longest-match search, preserving per-FEC label bindings across area boundaries.
Longest-match is defined as: the LDP FEC element exactly matches a RIB entry, OR the FEC element is a subset (more specific) of a RIB entry. A FEC element that is a superset of a RIB entry does NOT constitute a match.
The procedure is applicable to IPv4 and IPv6 prefix FEC elements, corresponding to address families 1 and 2 in the IANA 'Address Family Numbers' registry.
The Longest-Match Label Mapping procedure does not introduce any changes to the security considerations of RFC 5036 (LDP Specification). All LDP security properties remain unchanged.