Data Path Support

The support of SR-TE in the data path requires that the ingress LER push a label stack where each label represents a hop, a TE link, or a node, in the ERO for the LSP path computed by the router or the PCE. However, only the label and the outgoing interface to the first strict or loose hop in the ERO factor into the forwarding decision of the ingress LER. In other words, the SR-TE LSP only needs to track the reachability of the first strict or loose hop.

The first strict or loose hop of the SR-TE LSP is represented as an NHFLE to the SR shortest path tunnel. The rest of the SR-TE label stack can have a larger size and is modeled as another NHLFE referred to as a ‟super NHLFE”.

Therefore, an SR-TE LSP is modeled in the ingress LER data path as a hierarchical LSP with the super NHLFE tunneled over the NHLFE of the SR shortest path tunnel to the first strict or loose hop in the SR-TE LSP path ERO.

Some characteristics of this design are as follows.

The data path behavior at the LSR and egress LER for an SR-TE LSP is similar to that of the shortest path tunnel because there is no tunnel state in these nodes. The forwarding of the packet is based on processing the incoming label stack consisting of a node SID and/or adjacency SID label. If the ILM is for a node SID and multiple next hops exist, ECMP spraying is supported at the LSR.

The link-protect LFA backup next hop for an adjacency SID can be programmed at the ingress LER and LSR nodes (as explained in SR-TE LSP Protection).

A maximum of 12 labels, including all transport (including entropy), service, NGE, and OAM labels, can be pushed. The label stack size for the SR-TE LSP can be 1 to 11 labels, with a default value of 6.

The label stack size manipulation includes the following LER and LSR roles:

An example of the label stack pushed by the ingress LER and by an LSR acting as a PLR is illustrated in Figure: SR-TE LSP Label Stack Programming.

Figure: SR-TE LSP Label Stack Programming

On node A, the user configures an SR-TE LSP to node D with a list of explicit strict hops mapping to the adjacency SID of links A-B, B-C, and C-D.

Ingress LER A programs a super NHLFE consisting of the label for the adjacency over link C-D and points it to the already programmed NHLFE of the SR tunnel of its local adjacency over link A-B. The latter NHLFE has the top label and also the outgoing interface to send the packet to.

Note: The SR-TE LSP does not consume a separate backup super NHLFE; it only points the single super NHLFE to the NHLFE of the SR shortest path tunnel it is riding. When the latter activates its backup NHLFE, the SR-TE LSP will automatically forward over it.

LSR Node B already programmed the primary NHLFE for the adjacency SID over link C-D and has the ILM with label 1001 point to it. In addition, node B will preprogram the link-protect LFA backup next hop for link B-C and point the same ILM to it.

Note: There is no super NHLFE at node B because it only deals with the programming of the ILM and primary and backup NHLFE of its adjacency SIDs and its local and remote node SIDs.

VPRN service in node A forwards a packet to the VPN-IPv4 prefix X advertised by BGP peer D. Figure: SR-TE LSP Label Stack Programming shows the resulting data path at each node for the primary path and for the FRR backup path at LSR B.