[c-nsp] Replacing proper planning with dirty hacks (VLANextension over GRE & L2TP)

Brad Henshaw brad.henshaw at qcn.com.au
Tue Oct 30 17:00:51 EDT 2007


Rolf Mendelsohn wrote:

> What is the Radio network based on and what are the Specs.
> i.e. QinQ, Vlan's, MTU, PtP or PtMP (or a hybrid of the 2).

Hi Rolf,

The radio network is based on proprietary modifications to the
802.11g protocol as far as I'm aware. The topology is full mesh
- not just PtMP. One radio node may have visibility to multiple
other nodes and will select which next hop to use dynamically.

The radios are effectively routers and (somewhat strangely)
to the best of my knowledge they use BGP to make routing decisions
across the mesh, though this is abstracted from the user.
Connection from each radio to the wired network is typically an
access port. The wired and wireless interfaces on the radios
must be in different subnets. (as would be the requirement on
any router)

While the vendor released various software updates in an attempt
to implement 802.1q trunking from end to end, what was implemented
was purely cosmetic - i.e. the same VLAN IDs at each end, but
it wasn't possible to have the same subnets inside those VLANs
at different points on the mesh, because that couldn't be routed.
It looks as though the tags were being stripped by the radio before
being routed across the mesh, then re-applied at the remote end
before being regurgitated on the wire. (one must manually configure
a subnet to VLAN mapping on each radio to make this happen)

The vendor has since admitted that actual trunking or tunnelling
of 802.1Q-tagged frames across the mesh is not supported by the
units. (obviously if GRE, L2TPv3 or similar were to be used,
this is transparent to the radios)

Let me know if I need to clarify anything further.

Regards,
Brad


More information about the cisco-nsp mailing list