Figure 5-1 represents a basic internetwork containing MSDP speakers. Figure 5-1 is used to discuss the RPF-peer rules. With each rule, we provide a subsequent figure, based on Figure 5-1, to illustrate the specific situation.
All rules are examined from the perspective of router D. Router D has only two peers, which simplifies the explanation of the rules. The rules are applied to all MSDP peers that are in ESTABLISHED state at the time that the SA is received. Figure 5-1 is a generic diagram that does not include information about BGP peering sessions or other details on unicast routing that affect MSDP RPF-peer selection. These details are filled in as we discuss the rules.
The MSDP peering paths denoted in the figure are simply a chain of routers that provide MSDP
connectivity from router A to router B and from router A to router C. In this discussion, we are interested only in router D's RPF-peer decision, but keep in mind that each router must make its own independent decision.
Each rule can select only one peer, if any. It is important to note the first rule that matches is the RPF peer. Because of these two facts, the RPF-peer rules select only a single peer as the RPF peer.
5.4.1.1 RPF-Peer Rule #1: If the originating RP is a peer, it is the RPF peer
The first rule is simple. If router D is an MSDP peer with the originating RP, it uses the originating RP as its RPF peer. For example, if router B were the originating RP, it would be router D's RPF peer for all the SAs router B originates.
For the rest of the rules, we assume that router A is the originating RP of an SA message that router D just received from both router B and router C. Router D uses the rules described in the following sections to determine whether to accept the SA from router B or from router C.
5.4.1.2 RPF-Peer Rule #2: If the BGP next hop toward the originating RP is a peer, it is the RPF peer
If any peer is the BGP next hop of the active RPF route for router A's address, that peer is selected as the RPF peer. The BGP next-hop attribute is normally set to the address of BGP peer that advertised the route into the AS. This is not the case if the local AS uses a next-hop-self policy. If a next-hop-self policy is used, the BGP next-hop attribute is set to the address of the Internal Border Gateway Protocol
(IBGP) peer that received the route from a neighboring AS.
To illustrate this rule, in Figure 5-2, we add more detail about the BGP topology to the original figure depicted in Figure 5-1.
Assuming that the route for router A's address is learned through router B, rule 2 is fulfilled. The only question is "by which router?" If standard BGP policy is used, router B is chosen as the RPF peer because it is the BGP next hop. However, if router C implements a next-hop-self policy, router C is chosen as the RPF peer.
5.4.1.3 RPF-Peer Rule #3: If the BGP peer that advertised the route toward the originating RP is a peer, it is the RPF peer
At first glance, rule #3 looks identical to rule #2, but keep in mind that while routes learned from EBGP peers always have the BGP next hop set to the peer address, IBGP does not follow the same rules. An IBGP peer sets the next-hop attribute to be itself only if a next-hop-self policy is used. Additionally, if
route reflection or confederations are used, it is possible to receive a route from an IBGP peer that has
the next-hop attribute set to something other than that peer's address. To illustrate this rule, consider
Figure 5-3, in which the MSDP session between router B and router D is removed.
Figure 5-3. RPF-peer rule #3
This time router C does not implement a next-hop-self policy. Therefore the BGP next hop for the route to router A is router B's address and not router C's address. Because of this, rule #2 does not match any of router D's ESTABLISHED MSDP peers. Rule #3 is needed in order to accept SA messages originated by router A.
how rule #3 could be applied in a domain where route reflectors are used. Once that diagram is complete, point out an MSDP peering session that could be added to make rule #2 determine the choice of the RPF peer instead of rule #3.
Note
The original purpose of MSDP was to announce sources across multiple PIM-SM domains, but the protocol can also be used in the intradomain case. For example, the PIM's anycast RP mechanism relies on MSDP to announce sources between the multiple RPs in the domain. The main difference in the intradomain case is that the route to the originating RP is most likely not learned through BGP. Instead, it is learned via an IGP. If the IGP is a link-state protocol, either rule #1, static RPF peer, or mesh groups must be used. If the IGP is a distance-
vector protocol (DV protocol), rule #3 can be used. Specifically, the RPF peer is the
neighbor that advertised the route. This doesn't seem very attractive considering that the neighbors would have to be directly connected, and the primary goal of anycast RP is to provide load balancing and failover with routers in different topological locations.
5.4.1.4 RPF-Peer Rule #4: Of all the MSDP peers in the AS path toward the originating RP, the one with the highest IP address is the RPF peer
This rule is sort of a last-ditch effort to find some peer that is closer to the originating RP than the local router. In general, the design of an MSDP architecture should not rely on this rule. It is handy in a few situations, though. Looking back at Figure 5-1 and changing the underlying BGP topology, this rule is illustrated in Figure 5-4.
Figure 5-4. RPF-peer rule #4
In this case, router B and router C are router D's only MSDP peers. Furthermore, router B and router C cannot both be in the AS path for router D's route to router A. In Figure 5-4, let's assume that the path via router B is shorter and therefore is the AS path of router D's route to router A. In this case, router B is the only peer in the AS path toward router A, so it is selected as the RPF peer.
So how does router D know in which AS router B is located? After all, MSDP does not carry AS
information in any of its messages like BGP does. There are a few possible ways to determine in which AS each MSDP peer is located. One is to enable the user to specify the AS of each peer in the
configuration file of the router. Another is to have the router check the route to the peer and extract the information from the AS path of that route. The latter strategy is dangerous because the route to the peer may change and information may get out of synch. In the event router B and router C are in the same AS, the router with the highest IP address is selected as RPF peer as a final tiebreaker.
5.4.1.5 RPF-Peer Rule #5: If a static RPF peer is configured for the originating RP, it is the RPF peer
The fifth and final rule is a simple one. MSDP enables the configuration of static RPF peers. Each static RPF peer has a prefix range associated with it. That prefix range designates to which originating RPs the static RPF-peer configuration applies. If the prefix range is 0.0.0.0/0, the static RPF peer applies to all originating RP addresses and is called a default RPF peer.