BSCI question about : PIM Sparse mode and Sparse-Dense mode

xine xine · ‎06-01-2010

Hi

I'm currently preparing BSCI certification exam. I'm reading about multicast routing actually. I read the Authorized Self-Studdy Guide BSCI for my preparation to the exam.

First, I find something is a little bit in contracdition in the text of the author (base on my understanding… (I'm a French people, it make harder for me to undersand….)

In PIM Sparse mode the author say the source router register with the RP point. (Thats OK for me…) All receiving router should register with the RP or with source router base on the lowest cost to reach each other. After that all multicast data will be send to the RP and will be replicated to all receiving router from that point.

What is not clear to me is what about router who had register with the source router itsefl because they have a lowest cost to reach the source router rather then the RP ? They will receiving data from the RP or not ? If they receiving data from the RP why they had register with the source router in this case ?

Second question :

In PIM sparse-dense mode, how the automatic RP is work exactly in case of multi RP can be used, speacially if seperate RP for each source can be created ? The better RP for each source is the source router itself ?

"PIM sparse-dense mode allows the router to operate in sparse mode for sparse mode groups (those with known RPs) and in dense mode for other groups.

For maximum efficiency, multiple RPs can be implemented with each RP in an optimum location. Configuring, managing, and troubleshooting multiple RPs can be difficult if done manually. However, PIM sparse-dense mode supports automatic selection of RPs for each multicast source. For example, in Figure 9-19 Router A could be the RP for Source 1 and Router D could be the RP for Source 2.

Figure 9-19. Multiple RPs Are Supported with PIM Sparse Dense Mode

"

Base on my understanding router F is an better RP point for the "source 2" for all multicast router…. If receiver are also connected to router F for source 2 the trafic should not send to router D and back to router F after that …. This do not make sence to me….. in other way if the RP point should be closest to the receiver router C is an better RP for both source actually because we have only 2 receivers at this time !!

Someone can help me to understand ?

Thanks a lot !

Hitesh Vinzoda · ‎06-02-2010

Hi,

The Sparse-mode creates initially (*,G) shared tree entry in the routers where the receiver's are located. Once the multicast is received and if any shortest path is available to the source it will switchover and creates an (S,G) entry i.e Source tree at the receiver router, to avoid sub optimal routing of multicast packets from RP.

So intially they join the group to the RP and ultimately receives multicast from the source directly once switched over.

HTH

Hitesh Vinzoda

Please rate useful posts

Giuseppe Larosa · ‎06-02-2010

Hello Christien,

Once the source is known, multicast routers can check if it is better to join a source based tree (SPT bit) and to leave the shared tree that is rooted at the RP.

When a source PIM router registers with RP it means that it encapsulates the multicast packet inside a GRE packet with destination = RP address.

This is done because in PIM SM the delivery tree is unidirectional and branches can only exit from RP.

As the RP receives the GRE packet it extracts the multicast packet to be sent downstream to other routers or uses and it joins the source based tree that is rooted at the PIM router where the source is connected.

So only the sending router registers with the RP, routers with interested receivers join the shared tree rooted at the RP and they can switch to the source based tree as packets start to flow. (downstream routers send PIM joins towards the RP with appropriate flags signaling they want to join the shared tree)

The use of the RP allows to save resources (memory and cpu) on all routers by building a shared tree that is then optimized

There is a threshold for moving for SPT that has a default value so low that the move is done after receiving one packet.

There is a global command

ip pim spt-threshold

if infinite is provided as rate switch to SPT is disabled.

So traffic will go via the RP only if the RP is on the best path between source and receivers. likely for some receivers going via the RP is still the best choice even after considering the source but some receivers may receive traffic that has not gone via the RP (this depends on effective topology)

PIM sparse-dense mode solves a problem that Auto-RP has:

auto RP uses well known multicast addresses 224.0.1.39 and 224.0.1.40.

in order to route these packets that contain the RPs to groups mappings in PIM pure sparse mode an RP would be needed !

with pim sparse-dense mode routers use dense mode for groups not mapped to any RP and this solves the problem of auto-RP groups propagation.

To be noted that if instead of autoRP we use the bootstrap protocol that used the bootstrap message with destination 224.0.0.13 (PIM version2) this problem is not present

When using the so called anycast RP (that works with manual, autoRP or BSR) actually the RP used is the nearest one and all the routers acting as RPs need to communicate using MSDP for correct routing

Hope to help

Giuseppe

xine xine · ‎06-02-2010

Hi Giuslar,

Thanks a lot !

I do not understanding that part of your posted :

PIM sparse-dense mode solves a problem that Auto-RP has:

auto RP uses well known multicast addresses 224.0.1.39 and 224.0.1.40.

in order to route these packets that contain the RPs to groups mappings in PIM pure sparse mode an RP would be needed !

with pim sparse-dense mode routers use dense mode for groups not mapped to any RP and this solves the problem of auto-RP groups propagation.

To be noted that if instead of autoRP we use the bootstrap protocol that used the bootstrap message with destination 224.0.0.13 (PIM version2) this problem is not present

When using the so called anycast RP (that works with manual, autoRP or BSR) actually the RP used is the nearest one and all the routers acting as RPs need to communicate using MSDP for correct routing

Also, I don't you answer to my second question in your second part of your posted if it is I do not following you....

refering to my initial post :

"Base on my understanding router F is an better RP point for the "source 2" for all multicast router…. If receiver are also connected to router F for source 2 the trafic should not send to router D and back to router F after that …. This do not make sence to me….. in other way if the RP point should be closest to the receiver router C is an better RP for both source actually because we have only 2 receivers at this time !"

Thanks a lot !