Hi Weiguo,

Thanks, please see inline.

Thanks,
Ramki

From: Haoweiguo [mailto:[email protected]]
Sent: Sunday, August 31, 2014 8:31 PM
To: ramki Krishnan; [email protected]
Cc: [email protected]
Subject: 答复: [nvo3] triggering discussions on the multicast draft


Hi Ramki,

Pls see my comments inline with [weiguo].

Thanks

weiguo

________________________________
发件人: ramki Krishnan [[email protected]]
发送时间: 2014年8月31日 7:37
收件人: [email protected]<mailto:[email protected]>
抄送: 
[email protected]<mailto:[email protected]>
主题: [nvo3] triggering discussions on the multicast draft
Dear All,

We would like to kick start discussions on the multicast draft - 
https://datatracker.ietf.org/doc/draft-ghanwani-nvo3-app-mcast-framework/

From Section 3.2 “Replication at the source NVE”
>> When a VPN client has multiple multicast groups, [RFC6513]
     "Multicast VPN" combines all those multicast groups within each
     VPN client to one single multicast group in the MPLS (or VPN)
     core.  End result:  All messages from any multicast groups
     belonging to one VPN client will reach all the PE nodes of the
     client. I.e. any messages belonging to any multicast groups under
     Client XX will reach all PEs of the Client XX. When the Client XX
     only has a handful of PEs, there is not too much bandwidth wasted
     in the core.
One way to address this issue is to use replication at source NVE which will 
make sure that the multicast traffic will get delivered only to those NVEs 
where there are members; multicast groups don't get combined in any way. Are 
there other ways of addressing this issue? Comments on this topic or other 
comments on the draft would be much appreciated.

[weiguo]: Besides your suggested head-end replication at source NVE , there are 
other common solution can address the above problem.
1. Each multicast group in a VPN client maps to a single multicast group in the 
MPLS(or VPN) core, rather than all multicast groups in a VPN client mapping to 
a single public multicast group. This solution maybe rely on centralized 
controller to map(or allocate) a public multicast group for each private 
multicast group in a VPN client. Through SDN controller, the mapping between 
public and private multicast group can be consistent on each PE. Otherwise, if 
each PE allocates public multicast group for each private multicast group 
locally, different PEs cann't have same mapping result and will incur multicast 
traffic forwarding problem on core network.
[ramki] I guess the primary issue with this approach would be underlay 
multicast group scalability. Perhaps, this scheme can be used when the 
scalability needed is limited.

2. Centralized replication can be used to enhance efficiency and scalability of 
head-end replication at source PE. The solution is described in section3.3 in 
the draft.

Thanks,
Ramki
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3

Reply via email to