Multicast question

[shh5455]

Help me understand what I’m missing.  I understand that source trees use up more memory and shared trees use up less memory.  What I don’t understand is what is the point of using a shared tree if by default the leaf router switches over to the SPT (source-tree) immediately.

By default the SPT-threshold is set to 0.  Therefore, by default a leaf router will switch over to the SPT as soon as it receives a packet from the shared tree thus creating an (S,G) entry in its routing table.  Cisco seems to imply that the default setting is good enough for the majority of networks.  Seems like you’d end up having more entries – not only all of the (S,G) but all of the (*,G) also.  What am I missing?

Thanks

[Mahesh Gohil]

Hello,

You are correct but the only thing to consider is every three minutes source build tree again and if  there is no receiver

it will start pruning.

So it is wasted of bandwidth to span tree to those nodes where there no actual receivers

hope this helps

Regards

Mahesh

[shh5455]

Thanks for the reply. Does sparse mode have to be used with an RP (shared tree)? I believe we could use sparse mode with source trees as well.

Steven

[Mahesh Gohil]

Hello Steven,

sparse/dense are two protocol where dense believe in push model (flood and prune) while sparse use pull model.

sparse supports both model ...source and shared tree

whereas source/shared distribution tree is method of distributing traffic

hope this helps

Regards

Mahesh

[shh5455]

I understand that much, but it still doesn't answer my original question. Why use shared tree if it's just going to switch to source tree immediately?

Steven

[Mahesh Gohil]

Hello Steven,

As I already said..

Flood and prune process repeats every three minutes, till the time it prune again it will use extra resources which is

wastage of bandwidth.

consider a bigger network where there are around 500-700 routers how will you justify the this flood-prune mechanism

Hope this helps

Regards

Mahesh

[shh5455]

Ok, maybe we're talking about 2 different things. I thought you were talking about PIM-DM, but you may have been talking about DM-Fallback. I should add that we would want to use "no ip pim dm-fallback." So evaluate the 2 settings in light of the immediate SPT switchover and tell me what benefit the shared tree is:

Shared Tree

[Tharak Abraham]

Source based trees scale on an order of (SG * GN), and shared

trees scale on an order of (GN), where GN is the number of groups in the multicast domain and SG is the number of sources per group.

Thats a huge difference in reality !

eg: 10 sources with 10 Groups will create 10*10 or 100 States on the EVERY router in DM whereas the same will create just 10 States in an SM

Now coming to the SPT, thats an enhancement to the SM or making it an efficient way to stream.(considering the fact that the shortest path is the fastest)

Cisco by default do an SPT switchover immediately, but you can always change it based on the flow/BW

no ip pim dm-fallback on the other hand is to stop avoiding the use of DM, if the RP information is lost.

This is recommended in production

And again with DM, FYI..

Steady state operation and the state refresh does avoid the 3 minute flood and prune behaviour with PIMv2.

But still the router would have to maintain the state table of SG * GN