We are building a layer 2 WAN cloud to interconnect three sites (1 hub and 2 spokes).
Current setup of the three sites:
- Each site has 'unique' IP scheme (RFC 1918) and the sites are connected using site-to-site VPN.
We will be provided by a ethernet handoff at each site by the provider. The WAN provder has given the following directions:
- Each port needs to be configured as a trunk link (so that each site can send and recieve multiple VLANs)
- Each site is being assigned a unique VLAN.
- The provider does not support QinQ.
We dont want to extend VLAN's between sites. The main reason for getting the layer2 connection (versus layer3) was speed/price of the circuits.
We just want to interconnect the three sites (in a full mesh) and run OSPF for interconnectivity so that we can use the S2S VPN tunnels as a backup.
Couple of questions on the design:
- If each site is going to have a trunk link to the cloud, how can I ensure that my spanning-tree is not spanned across the sites (i.e. how can I NOT extend my spanning-tree domain to other sites).
- If each site will be assigned a unique VLAN (by the provider) and since I am planning to route traffic between the sites using OSPF (instead of switching traffic), I believe on EACH site, I will need to define an SVI (switched virtual interface) for the provider assigned VLAN for not just the local site but also for the remote sites' VLAN:
Example if Hub site is assigned VLAN3, Remote-site1 is assigned VLAN1 and Remote-site2 is assigned VLAN2 then on each switch which terminates the layer2 handoff I will need to define an SVI for all three VLAN's and include these interfaces in OSPF and advertise the local subnets of each site in OSPF.
- Is there any other design options giving this scenario. The provider is not too helpful.
Thanks!
