Why PIM-Register mechanism is required?

[dhananjay95929]

I am not able to grasp my head around the concept of why PIM Register mechanism is required.

Everyone seems to keep repeating the process of how PIM Register works instead of explaining why its in place.

As I understand it, RP is required to discover sources so that burden of locating sources is centralized as opposed to being distributed to each PIM router.

Why does the DR of the source's segment need to register with the RP and then once registration is complete the tree switches from RPT/Shared Tree to Source Tree?

[Joseph W. Doherty · Accepted solution]

"As I understand it, RP is required to discover sources so that burden of locating sources is centralized as opposed to being distributed to each PIM router."

More or less. And the RP, I recall, doesn't discover routers with multicast sources, they register to the RP and routers with receivers also go to the RP. I.e. the RP allows a very quick "hook up" between multicast sources and multicast receivers. The "why" is to improve efficiency and performance for making this logical connection.

"Why does the DR of the source's segment need to register with the RP and then once registration is complete the tree switches from RPT/Shared Tree to Source Tree?"

BTW, you can turn off moving to the SPT, but the reason you generally want that is to also improve performance and avoiding loading down additional devices that need not be involved with the multicast flow from source to receiver.

Basically, SM mode increases complexity on the routers, but it offers many benefits for supporting multicast flows, especially as the topology grows in size.

[dhananjay95929]

Thanks, that kind of makes sense. So the PIM register process is required so that RP can know about the source of traffic because there is no other automatic discovery of sources from the RP perspective? (kind of odd and weird because from downstream router perspective there exists a method to discover RP - Cisco's AutoRP, static RP and BSR so why not have a similar mechanism for RP to discover sources?).

So once the RP knows about the source of the traffic it then sends this source IP address information downstream towards the DR of the receiver's segment which switches to the Source Path Tree (S,G) from the RPT (*,G) once it gets this IP address S of the source of traffic?

[Joseph W. Doherty]

Well you do have a similar mechanism, the multicast source routers notifying the RP router. The other routers "know" the RP IP, so instead of the RP device trying to discover multicast sources, the inverse happens. Result, though, is the RP "knows" about multicast sources.

This is the PIM-SM approach, but its not the only multicast routing approach, just a fairly nice/good one.

Yes, to your second question, but also keep in mind, all the other multicast routers that may be involved with a migration to SPT, and what's involved to make that switch.

[Jon Marshall]

Just to clarify the RP does not send the IP address information to any router. 

It simply forwards on the multicast packets on the shared tree and the router closest to the receiver then knows the source IP from the received packets. 

Jon

[dhananjay95929]

Thanks for that information, makes sense, so the RP simply forwards/steers the original IP packet from the source through the PIM tunnel towards the next PIM neighbor downstream keeping the source IP address within the packet unchanged. The downstream PIM routers that are neighbors with each other and the RP follow the same procedure keeping the source IP address within the packet unchanged until the packet reaches the Last Hop Router/DR of the receiver. When the packet reaches the LHR it knows its the last hop to the receiver(s) and can use the source IP address within the packet to determine the source of traffic and can switch to the SPT if desired.

[paul driver]

Hello

I know there is a lot more to it but in simplistic 10,000 ft perspective- My understanding:

RPT - shared
1) source to RP
2) RP to client <-----(not probably shortest path but source / client is now known)

SPT -  shortest
Source to Client <--Is the shortest path ( could be via RP anyway)

res
Paul

[Joseph W. Doherty]

I don't recall whether the LHR is the one that makes the decision to switch to SPT from source to receiver. Remember, other routers can be impacted.

Before you (optionally) switch to STP, the source "sends" to the RP via a STP to it, and the the RP "sends" to the final receiver via STP to it. Which, also as Paul notes, the path through the RP might actually be the STP from source to receiver.

If there is a change in path for source to receiver STP, I recall (?) it's possible for the receiver (or at least the LHR) to receive two copies of the multicast flow's packets during the transition.