P2MP RSVP LSP is specified in RFC 4875, Extensions to Resource Reservation Protocol - Traffic Engineering (RSVP-TE) for Point-to-Multipoint TE Label Switched Paths (LSPs).
A P2MP LSP is modeled as a set of source-to-leaf (S2L) sub-LSPs. The source or root, for example the head-end node, triggers signaling using one or multiple path messages. A path message can contain the signaling information for one or more S2L sub-LSPs. The leaf sub-LSP paths are merged at branching points.
A P2MP LSP is identified by the combination of <P2MP ID, tunnel ID, extended tunnel ID> part of the P2MP session object, and <tunnel sender address, LSP ID> fields in the P2MP sender_template object.
A specific sub-LSP is identified by the <S2L sub-LSP destination address> part of the S2L_SUB_LSP object and an ERO and secondary ERO (SERO) objects.
The following are characteristics of this feature:
Supports the de-aggregated method for signaling the P2MP RSVP LSP. Each root to leaf is modeled as a P2P LSP in the RSVP control plane. Only data plane merges the paths of the packets.
Each S2L sub-LSP is signaled in a separate path message. Each leaf node responds with its own resv message. A branch LSR node forwards the path message of each S2L sub-LSP to the downstream LSR without replicating it. It also forwards the resv message of each S2L sub-LSP to the upstream LSR without merging it with the resv messages of other S2L sub-LSPs of the same P2MP LSP. The same is done for subsequent refreshes of the path and resv states.
The node drops aggregated RSVP messages on the receive side if originated by another vendor’s implementation.
The user configures a P2MP LSP by specifying the optional create-time parameter p2mp-lsp following the LSP name. Next, the user creates a primary P2MP instance using the keyword primary-p2mp-instance. Then a path name of each S2L sub-LSP must added to the P2MP instance using the keyword s2l-path. The paths can be empty paths or can specify a list of explicit hops. The path name must exist and must have been defined in the config>router>mpls>path context.
The same path name can be re-used by more than one S2L of the primary P2MP instance. However the to keyword must have a unique argument per S2L as it corresponds to the address of the egress LER node.
The user can configure a secondary instance of the P2MP LSP to backup the primary one. In this case, the user enters the name of the secondary P2MP LSP instance under the same LSP name. One or more secondary instances can be created. The trigger for the head-end node to switch the path of the LSP from the primary P2MP instance to the secondary P2MP instance is to be determined. This could be based on the number of leaf LSPs which went down at any specific time.
The following parameters can be used with a P2MP LSP: adaptive, cspf, exclude, fast-reroute, from, hop-limit, include, metric, retry-limit, retry-timer, resignal-timer.
The following parameters cannot be used with a P2MP LSP: adspec, primary, secondary, to.
The node ingress LER does not inset an adspec object in the path message of an S2L sub-LSP. If received in the resv message, it is dropped. The operational MTU of an S2L path is derived from the MTU of the outgoing interface of that S2L path.
The to parameter is not available at the LSP level but at the path level of each S2L sub-LSP of the primary or secondary instance of this P2MP LSP.
The hold-timer configured in the config>router>mpls>hold-timer context applies when signaling or re-signaling an individual S2L sub-LSP path. It does not apply when the entire tree is signaled or re-signaled.
The head-end node can add and, or remove a S2L sub-LSP of a specific leaf node without impacting forwarding over the already established S2L sub-LSPs of this P2MP LSP and without re-signaling them.
The head-end node performs a make-before break (MBB) on an individual S2L path of a primary P2MP instance whenever it applies the FRR global revertive procedures to this path. If CSPF finds a new path, RSVP signals this S2L path with the same LSP-ID as the existing path.
All other configuration changes, such as adaptive/no-adaptive (when an MBB is in progress), metric-type te, no-frr, path-computation-method/no path-computation-method, result in the tear-down and re-try of all affected S2L paths.
MPLS requests CSPF to re-compute the whole set of S2L paths of a specific active P2MP instance each time the P2MP re-signal timer expires. The P2MP re-signal timer is configured separately from the P2P LSP. MPLS performs a global MBB and moves each S2L sub-LSP in the instance into its new path using a new P2MP LSP ID if the global MBB is successful. This is regardless of the cost of the new S2L path.
MPLS requests CSPF to re-compute the whole set of S2L paths of a specific active P2MP instance each time the user performs a manual re-signal of the P2MP instance. MPLS then always performs a global MBB and moves each S2L sub-LSP in the instance into its new path using a new P2MP LSP ID if the global MBB is successful. This is regardless of the cost of the new S2L path. The user executes a manual re-signal of the P2MP LSP instance using the command: tools>perform>router>mpls>resignal p2mp-lsp lsp-name p2mp-instance instance-name.
When performing global MBB, MPLS runs a separate MBB on each S2L in the P2MP LSP instance. If an S2L MBB does not succeed the first time, MPLS retries the S2L using the re-try timer and re-try count values inherited from P2MP LSP configuration. However, there is a global MBB timer set to 600 seconds and which is not configurable. If the global MBB succeeds, for example, all S2L MBBs have succeeded, before the global timer expires, MPLS moves all S2L sub-LSPs into their new path. Otherwise when this timer expires, MPLS checks if all S2L paths have at least tried once. If so, it then aborts the global MBB. If not, it continues until all S2Ls have re-tried once and then aborts the global MBB. After global MBB is aborted, MPLS moves all S2L sub-LSPs into the new paths only if the set of S2Ls with a new path found is a superset of the S2Ls which have a current path which is up.
While make-before break is being performed on individual S2L sub-LSP paths, the P2MP LSP continues forwarding packets on S2L sub-LSP paths which are not being re-optimized and on the older S2L sub-LSP paths for which make-before-break operation was not successful. MBB therefore results in duplication of packets until the old path is torn down.
The MPLS data path of an LSR node, branch LSR node, and bud LSR node is able to re-merge S2L sub-LSP paths of the same P2MP LSP in case their ILM is on different incoming interfaces and their NHLFE is on the same or different outgoing interfaces. This could occur anytime there are equal cost paths through this node for the S2L sub-LSPs of this P2MP LSP.
Link-protect FRR bypass using P2P LSPs is supported. In link protect, the PLR protecting an interface to a branch LSR only makes use of a single P2P bypass LSP to protect all S2L sub-LSPs traversing the protected interface.
Refresh reduction on RSVP interface and on P2P bypass LSP protecting one or more S2L sub-LSPs.
A manual bypass LSP cannot be used for protecting S2L paths of a P2MP LSP.
The following MPLS features do operate with P2MP LSP:
BFD on RSVP interface
MD5 on RSVP interface
IGP metric and TE metric for computing the path of the P2MP LSP with CSPF
SRLG constraint for computing the path of the P2MP LSP with CSPF. SRLG is supported on FRR backup path only
TE graceful shutdown
Admin group constraint
The following MPLS features are not operable with P2MP LSP:
Class based forwarding over P2MP RSVP LSP
LDP-over-RSVP where the RSVP LSP is a P2MP LSP
DiffServ TE
Soft preemption of RSVP P2MP LSP