In the previous lessons, you saw how PIM Sparse Mode, SSM, and Bidirectional PIM build multicast trees inside a single network.
But every one of them assumes the source and the receiver share the same PIM domain.
What happens when they do not?A Receiver in Another Domain
Imagine your network spans two autonomous systems.
AS 100 contains the multicast Server. AS 200 contains PC1, the receiver.
Each AS runs its own PIM domain, with its own Rendezvous Point.R1 is the RP for AS 100.
R2 is the RP for AS 200.

Figure 1 – Inter-domain topology with two PIM domains
Imagine PC1 wants to receive group 239.1.1.1, so it sends an IGMP Membership Report to its last-hop router.
Inside AS 200, R2 (the RP) now knows that PC1 is interested.
Figure 2 – PC1 sends an IGMP Join to R2 in AS 200
But R2 has no idea who can send to that group.
There is no source registered in AS 200.Answer the question below
R2 knows PC1 wants the group, but what is still missing in AS 200?
A Source in Another Domain
Meanwhile, on the other side of the WAN link, the Server starts sending the multicast stream.
Just like in PIM Sparse Mode, the first-hop router in AS 100 registers the source with its local RP.
Figure 3 – R1 learns about the source in AS 100
R1 now has a registered source for group 239.1.1.1.
Inside AS 100, everything works exactly as PIM-SM expects.
But there is no receiver in AS 100. PC1 is on the other side, behind R2.Answer the question below
R1 has registered the source. What is missing in AS 100?
Two Domains, Two Blind Spots
So you end up with two RPs, each stuck with half of the puzzle.

Figure 4 – The two RPs are isolated
Look at what each RP is thinking.
R1 has the source but nobody asking for it.
R2 has a receiver but no source to pull from.
So why can't they just talk to each other through PIM?
Because PIM-SM does not cross domains.
PIM Register, PIM Join, and the RP itself are all scoped to a single domain.
There is no native mechanism for one RP to learn about sources in another domain.What you need is a way for R1 to tell R2: “Hey, I have a source for group 239.1.1.1.”
That is exactly the job of MSDP.Answer the question below
Does PIM Sparse Mode signaling cross AS boundaries?
Time to meet the protocol that solves this.
Multicast Source Discovery Protocol (RFC 3618) solves this exact problem.
It does not transport multicast traffic. It only carries source announcements between RPs.The MSDP Peering
MSDP works as a peering between RPs.
Two RPs in different domains open a TCP session and exchange information about active sources.
Figure 5 – MSDP peering between R1 and R2
The peering runs over TCP port 639.
Once established, the session stays up and exchanges keepalives every 60 seconds by default.
Notice what MSDP does not do.
It does not replace PIM. It does not carry multicast packets. It does not merge the two PIM domains.Each domain still runs its own PIM-SM independently.
MSDP sits on top, carrying announcements only.Answer the question below
Does MSDP carry multicast packets between the two domains?
The Source Active Message
When R1 learns about a new active source in its domain, it sends a Source Active (SA) message to every MSDP peer.
The SA message is the heart of MSDP.
Figure 6 – R1 announces the source to R2 with an SA message
Each SA message carries three pieces of information:
The source IP address (10.1.1.10)
The group address (239.1.1.1)
The originating RP (R1)
When R2 receives the SA, two things happen.
First, R2 stores the (S, G) pair in its SA cache.
Second, R2 checks whether anyone in AS 200 is interested in that group.Remember PC1 from Figure 2?
PC1 already sent an IGMP Join for 239.1.1.1, so R2 has a receiver waiting.
Now R2 has both halves of the puzzle: a receiver locally, and a source it just learned about through MSDP.Answer the question below
Which TCP port does the MSDP peering use?
MSDP told R2 that a source exists. But announcements alone do not move multicast packets.
The next step is to build a forwarding path between the source and the receiver.The Cross-Domain (S, G) Join
Now that R2 knows the source IP, it does not need to go through any RP to reach it.
R2 sends a PIM Join (S, G) toward 10.1.1.10.
This is the same Join mechanism used inside any PIM-SM domain.
But this time, it crosses the AS boundary because R2 knows exactly where the source lives.
Figure 7 – R2 sends a PIM Join (S, G) toward the source
The Join follows R2's unicast routing table, hop by hop, all the way to R1.
Each router along the path installs an (S, G) entry for the multicast tree.Notice what just happened across the AS boundary.
MSDP carried the announcement. PIM carried the Join.
Each protocol does its own job.Answer the question below
Which protocol carries the cross-domain Join from R2 toward the source?
End-to-End Multicast Delivery
Once the (S, G) tree is built, the multicast stream finally has a path from end to end.
Traffic flows from the Server, through R1, across the WAN link, through R2, and down to PC1.
Figure 8 – Multicast traffic crosses both domains
The traffic crosses the AS boundary as native multicast.
There is no tunnel, no encapsulation, no relay through MSDP.
MSDP did its job at the start, then stepped out of the data path entirely.Think of MSDP as a phone call between two RPs.
It is how they tell each other what they have.
The actual delivery is still done by PIM, in each domain, on its own.Answer the question below
Does MSDP carry the multicast traffic itself between domains?
Answer the question below
What type of message does an RP send to announce an active source to its MSDP peers?
Answer the question below
R2 learns SAs for three groups, but receivers joined only two: for how many (S, G) does it send a Join?