I'm having some trouble understanding how to operate VPLS and TE tunnels in parallel between a pair of routers.
I'd like to operate multiple TE and VPLS tunnels in parallel between a pair of routers. Right now when creating a VPLS tunnel, a pair of TE tunnels is automatically generated for bi-directional transport. If ever there are any existing TE tunnels on the head and/or tail-end router that are matching the transport address of the respective VPLS tunnel, those will be used instead.
That means that all the bandwidth constraints in-place for those TE tunnels, automatically also applies to the VPLS tunnels that are running on top. Since the TE tunnel pair is selected automatically during initialization phase it's somehow impossible to predict which TE tunnels are going to be used used for transport.
My question: How to operate multiple VPLS-LDP tunnels in parallel to RSVP-TE tunnels between a pair of routers w/o having the VPLS tunnels being randomly placed on any of the existing TE tunnels?
During my research I used to came across a post where someone suggested to just use a different set of loopback addresses for the TE tunnels as well as the respective VPLS tunnel to 'steer' the tunnel onto a specific pair. I've tried this but even the TE tunnels wouldn't come up. The loopback addresses are configured and are being advertised using OSPF within the backbone area.
Of course, using one set of loopback addresses per VPLS tunnel would produce a massive overhead in regards to IP address management. However, as long as "preferred path selection" wasn't implemented with ROS I don't see any way to make it work). It was confirmed by the MT support that preferred path selection wasn't implemented yet (!).
I'm attaching a diagram that hopefully helps you guys understanding what is outlined above.
I consider this a valid use-case for an MPLS domain to run multiple VPLS and/or TE tunnels in parellel between two routers.
Now I'd like to know how you guys are dealing with this? It's a valid use case which IMHO renders the platform useless if it can't be accomplished. However, if you think that running multiple L3+L2 tunnels in parallel between two routers isn't a valid use-case at all please let me know.