[j-nsp] Trunk port assignment

[Nils Swart]

Hi Bala,

>>  From your information it seems to me you're trying to transport a set
>> of VLANs on this gigabit ethernet interface to the other side of your
>> network. On the core-facing interface of this RSTN box, what are you
>> using ? i.e. MPLS based transport (L2VPN or VPLS) or VLAN stacking ?
>
> [Bala 03/29] The core is configured for VPLS transport. You are right, I
> am transporting a set of VLANs from the Riverstone PE to the 'other'
> side of my core. Granted, JNPR and RSTN both don't 'talk' the same VPLS
> language (support diff drafts) but I was wondering what would happen if
> the 'other' side were a Juniper M. Or, how to transport a set of VLANs
> between two Juniper M using VPLS core (specifically BGP-VPLS - one that
> JunOS supports as we speak)

Generally speaking the VPLS implementation of both RSTN and JNPR differ 
indeed, but you'll be able to stitch both together with one downside: 
there will not be any active signalling or autoprovisioning.

On the JNPR side you can terminate a standard pseudowire in a VPLS 
instance, allowing a remote switch (MTU-s in H-VPLS talk) to participate 
in a VPLS domain without having to have a full mesh of tunnels to each 
other participating PE.

As far as transporting multiple VLANs in a multipoint-to-multipoint 
fashion (p-2-mp would be spokes that cannot communicate to eachother but 
can only communicate with the hub, achievable by working with vrf targets 
just like any l3vpn), a VPLS domain 'switches' on MAC address only, and as 
such will not keep track of which tag and mac address share the same 
destination PE.

In the case where multiple VLANs have to be transported over a common MPLS 
core, i would see whether or not there is a need for mp-2-mp connectivity 
and build vpls instances per vlan for those cases. And in other scenarios 
one could transparantly move a full pipe from one side to the other using 
pseudowires.

>> Translational Cross Connects and VPLS are described in that guide as well.
> [Bala 03/29] Yes, I did refer to this too. But these seem "point to
> point". What abt a "point to multipoint" solution where I want to
> transport all vlans across an VPLS enabled core ?

-Nils