MSDP Flood Reduction
Last updated
Last updated
Load inter-as.multicast.msdp.configured.cfg
MSDP peering is already configured. There is a partial mesh of MSDP peerings. CSR9 and XR2 are not MSDP peers.
Currently, CSR9 will reflect MSDP SAs received from CSR10 to XR1 and vice versa. This also happens on XR2. Find a way to eliminate this redundant flooding on these two routers.
MSDP mesh groups sort of implement reverse BGP RR functionality. All MSDP peers in the same mesh group are treated as in regular iBGP, in which an SA from one peer is not flooded to any other peers in the mesh group. The idea is that all peers are fully meshed, so any flooding would be redundant.
MSDP mesh groups are typically used within an AS for Anycast RP, however we can use it in this inter-domain lab to reduce flooding.
MSDP mesh groups are also useful for turning off the RPF check. The RPF check for each peer in a mesh group is completely disabled. This is allowed because an SA from this peer will not be reflected to any other peers in the mesh-group, therefore the loop prevention is forced in the behavior of update propagation, just like iBGP split horizon.
All SAs from mesh group peers are still reflected to any peers that are not in the same mesh group.
Source traffic from CSR2 and notice that CSR9 does not reflect the SA message to XR1.
Note that strangely CSR9 will show no “accepted-SAs” from CSR10 even though the SA is in the cache. My guess is that this is because RPF checks are completely disabled for this peer, and the “accepted-SAs” must reflect SAs that passed an explicit RPF check.
We can see the same thing on XR2 but using a slightly different method. Examining MSDP stats we can see that no SAs have been sent to either peer.
