Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Announcements

Attention: The Community will be in read-only mode on 12/14/2017 from 12:00 am pacific to 11:30 am.

During this time you will only be able to see content. Other interactions such as posting, replying to questions, or marking content as helpful will be disabled for few hours.

We apologize for the inconvenience while we perform important updates to the Community.

New Member

Multicast VPN General Design Questions

All,

I have several general questions about Multicast VPN.  I have read conflicting information about SSM and the need for RPs within the Service Provider backbone.  To my understanding, SSM does not need a RP and per most documents I read, to enable multicast to transverse from CE1 to CE2, on the VRF terminating that customers, I create the MDT utilizing the same group address somewhere in the 232.0.0.0/8 family on both PEs terminating the CEs.  On the PEs, in particular the interfaces facing the SP backbone or P routers, I enable sparse-mode on all interfaces throughout the backbone.  I also enable ip pim ssm default on all SP devices.

Speaking to the PE side facing the CE, I either turn on sparse-dense mode or possibly define a RP for that VRF using the ip pim vrf X.X.X.X to the customer's RP.

To summarize the steps:

1) define the MDT under the VRF which must match on all PE devices for that VRF

2) enable Sparse Mode on all Service Provider backbone devices

3) enable ip pim ssm default on all SP devices

The questions are, do I need to enable IGMPv3 on all SP backbone links and do I need to create a RP within the SP backbone?  Per documentation I have read, a RP is not required with SSM but then again, I have seen some discussions where people were advised to turn up a RP.

This is not a provider to provider situation.  This within the same AS.

7 REPLIES
Cisco Employee

Re: Multicast VPN General Design Questions

Hi,

Your understanding is right that you dont need RP while PIM mode is SSM. PIM SSM works a bit siilar as PIm sparse-mode. But instead of sending the PIM join towards RP, it should be sent towards the Source. So you need to know the Source from whom the receiver is interested to receive the multicast stream.

Your summarised steps is partially fine to bring MVPN up.

When SSM is enabled on a multicast domain, you need some way to know the multicast group to join and the source from which you need the Multicast stream.

On MVPN scenario with SSM on SP backbone, you need to know the MDT group and the source address of all PE routers whch are enabled with the same MDT group. This can be done by enabling MDT SAFI in BGP.

So the basic requirement when PIM SSM is enabled on a MVPN scenario is that you need to enable MDT SAFI in BGP as below,

router bgp
address-family ipv4 mdt
neighbor activate

Now if you check "show ip bgp ipv4 mdt" you can see the other PE's source address and the MDT group address using which the local device will send PIM join and join the stream to get the PIM control packets.

I have included few more points in your summarised stpes to make it clear,

1) define the MDT under the VRF which must match on all PE devices  for that VRF

2) enable Sparse Mode on all Service Provider  backbone devices

3) enable ip pim ssm default on all SP devices (Be aware if you configure "ip pim ssm default" you can only use 232.0.0.0/8 range)

4) Enable bgp MDT SAFI in all PE routers.

5) Now VRF PIM neighborship over the MDT tunnel interface should be established.

HTH,
Nagendra

New Member

Re: Multicast VPN General Design Questions

First, thanks for your reply.  We have all of the configuration completed and are still having issues getting it to flow across but I believe the issue is on the CE side or what would be considered the CE side.  Essentially we are performing the layer 3 routing for the customer.  There is nothing but a layer two switch at each CE site.  We have a CE/PE.  The PE then connects to the P which then connects to the other PE/CE.  I am guessing that the issue is that either a RP needs to be defined at the VRF as in "ip pim vrf TEST X.X.X.X" or on the interface facing the customer layer two switch, we switch from sparse mode to sparse-dense mode.

Silver

Re: Multicast VPN General Design Questions

Hi

SP domain is normally broken into parts one is customer facing and another is core facing. In case of MVPN you can enable SSM for SP domain but doesn't require IGMPv3 because SP domain is only responsible for forwarding of customer traffic.

If you are planning to multicast for enterprise, in that case you require IGMPv3 if SSM is enabled in the domain. Now the question comes why IGMPv3 is required in SSM, because in SSM, every equipment doesn't know about the RP address, so IGMPv3 provides the source information at first hop.

In case of SSM, cleary RP is not required because no *,G entries created in it only S,G entries will stay in boxes.

For more on SP MPLS domain please check http://www.mplsvpn.info/2009/05/deploying-testing-of-ssm-in-service.html

regards

Shivlu Jain

http://www.mpslvpn.info

New Member

Re: Multicast VPN General Design Questions

First, thanks for your reply.  We have all of the configuration completed and are still having issues getting it to flow across but I believe the issue is on the CE side or what would be considered the CE side.  Essentially we are performing the layer 3 routing for the customer.  There is nothing but a layer two switch at each CE site.  We have a CE/PE.  The PE then connects to the P which then connects to the other PE/CE.  I am guessing that the issue is that either a RP needs to be defined at the VRF as in "ip pim vrf TEST X.X.X.X" or on the interface facing the customer layer two switch, we switch from sparse mode to sparse-dense mode.

Silver

Re: Multicast VPN General Design Questions

In this case you need to define a RP under vrf and that RP should belong to your vrf domain. Remember it it should be part of your SP domain. When ever the traffic comes from CE, PE will encapsulate it and send over SSM.

regards

Shivlu Jain

http://www.mplsvpn.info              

Re: Multicast VPN General Design Questions

Jason,

The provider Multicast traffic is totally independant of the customer multicast traffic. The Provider is fully transparent to the Customer Multicast Traffic.

Here is the Concept:  (( CEs signal the use of Data-mdt with the provider PEs via udp port 3232 , THe solution creates a tunnel interface between the PEs. Unlike GRE, this tunnel interface is not a point to point interface, PEs uses this tunnel to unicast the Multicst packet to the remote PEs, a Unique Source at the provider edge is required, this source recommended to be the Loopback interface of the BGP session, a Unique Group is also required. In current Cisco's Implementation of Multicast VPN, Pim Bidirectional is not supported and will be implemented once it proves stability. Pim sparse-dense Mode on all PE interfaces facing CE interfaces has to be enabled, this allows for carrying all customer multicast whether RP information, dense or Bootstrab messages. all Pim Signaling between PEs are carried by MDT.))

So to Conclude, running Mvpn on the service provider Network needs the following:

1- Pim Spars Mode on the Core.

2- Pim Sparse Dense on the Provider Edge facing CEs.

2- a Unique MDT-DATA Group per VRF.

3- a Unique Source per each VRF.

HTH

Mohamed

Silver

Re: Multicast VPN General Design Questions

Please check the given link and download the document. That will clear you lot of things.

http://www.mplsvpn.info/2009/05/deploying-testing-of-ssm-in-service.html

regards

Shivlu Jain

661
Views
0
Helpful
7
Replies
CreatePlease to create content