Question on Multicast VPN

[chintan-shah]

Hi,

In service provider to provide Multicast VPN to customer, they run PIM-SM or PIM-SSM. PIM-SM for default MDT.

PIM-SM require RP and i learnt that best to use Anycast RP.

Now, I have 13 locations having P route connected to form core. Do i need to run Anycast RP on each core local to location and run MSDP between them to exchange SA?

What should be best solution to support PIM-SM in Core for Multicast VPN on MPLS network ?

I guess, Custoemr RP,MSDP would be under VRF instance for each customer.

Please correct me if i am wrong , can suggest any good document.

Regards,

Chintan

[Harold Ritter]

Chintan,

That is correct. If the same data MDT group address is reused for let's say 2 C-Domain multicast stream and that two PEs need to subscribe to one of each, then both PEs will receive both C-Domain multicast streams. This is not optimal but it is way better than staying on the default MDT ;-)

Regards

[chintan-shah]

Hi Hritter,

As Multicast VPN from service provider perspective is encapsulating C-packet in to P-Packet with int SP network and that is using GRE.

I assume that there will be overhead of 24 byte of GRE and we will need to change IP MTU from 1500 to 1524 on all P and PE router so that C-Packet with 1500 byte ( with DF bit set) can be send across to other site from SP network.

Now question I have is, in our current network we have "mtu 1512" ( i.e. MTU 1512, MPLS MTU 1512, CLNS 1509 (for IS-IS) ) and "ip mtu 1500".

If we change "ip mtu 1524" , i belive MPLS MTU not require to change but still physical MTU require to change so what should be correct MTU for MPLS, CLNS, IP in our network now to take care of Multicast VPN and normal unicast traffic ??

Regards,

[shivlu jain]

hi chintan

MVPN doesnot require the MPLS. MVPN solution is proposed by rosen and rahul aggarwal because mpls doesnot understand the multicast and its very difficult to carry the multicast traffic from mpls core. For that rosen proposed that dynamic gre tunnel will be built where there will default mdt. So gre tunnel is having mtu of 1514 and it simply works as ip packet. So there is no change required in the core.

regards

shivlu jain

[chintan-shah]

Hi Shivlu,

Yes, MPLS doesn't reqire MPLS and that's reason as of now GRE we need to use with unique Service provider Multicast group to allow customer with overlapping Multicast group, they have.

But, as you know this dynamic GRE tunnel will have overhead and by default GRE overheald is 24 byte. I learnt from Cisco that IP MTU should be 1524 not 1514 to allow 1500 byte payload in C-Packet. We need confirmation from Hritter on this overheald 14 or 24 byte :-).

But in any case, we have IP MTU 1500 on our PE and P router interface with in core network. and we configure MTU 1512 which in turn automatically derive MPLS MTU 1512, CLNS MTU 1509 ( MTU-3) for IS-IS. So now if at all I need to change the IP MTU to 1514 or 1524, i also need to change physical MTU and thus automatically other derived MTU ( MPLS, CLNS) gets changed.

So, I understood from draft-rosen that it is better to fragrment the Multicast packet before it gets encapsulated such that you havce Max 1476( 1500-24) or 1486 ( 1500-14) IP paylod from customer, which can be configure with in VRF : "mdt mtu ". Only problem with this is if c-packet with DF bit set , it will get drop..

So, What should be correct option to take care of this MTU business :-) ??

Regards,

[shivlu jain]

Chintan

If you run show ip pim mdt command on PE and get to know about the MTI tunnel. There after run show int tunnel x; In that tunnel it shows 1514 bytes.

But I think hritter can explain more better because of week end we will get the reply by tomorrow.

regards

shivlu jain