L2TPv3 Ethernet Pseudowire Scalability

[David_Mitchell]

Good afternoon all,

I have labbed a L2TPv3 Ethernet Pseudowire topology to try and learn more about the differences between L2TPv3 vs L2 VPN o/MPLS etc and now that I have it labbed and working I can see how it would scale from a SP point of view i.e. adding additional customers would be easy enough, but what If an existing customer using the pseudowire wanted to add another site to the mix off the same PE router?

Or in the example I have used, what if there were another PE router elsewhere for the new customer which will facilitate the xconnect peering but you would have to change the IP addressing between the existing PE and CEs to make the OSPF customer network an OSPF Broadcast network. 

L2TPv3 Pseudowire doesn't seem scalable at all compared to L2 VPN over MPLS using vrfs?

Any clear explanaitions welcome

I have attached my topology for your conveniece

Thanks in advance

Mitchell

[Jason Fraioli]

David, this is the same problem I am currently facing.  With multipoint L2TPv3, you need to provision additional interfaces, on the LAN side of the router, to create the xconnect mesh between all destinations.

R1 (E0/0) ------xconnect to------ R2 Loopback IP with unique VC ID (12, R1 to R2)

R1 (E0/1) ------xconnect to------ R3 Loopback IP with unique VC ID (13, R1 to R3)

R2 (E0/0) ------xconnect to------ R3 Loopback IP with unique VC ID (23, R2 to R3)

From the perspective of R1, you would connect E0/0 and E0/1 into a switch and let spanning tree take over.  The same would be true at the other two routers.  I’ve got this working with a single vlan right now.  You may be able to do this with sub interfaces as well.  I don’t have a lab to test with at the moment.  Like I said though, I am running into the same scaling problems.