The MPLS WG Archive

[Jim Guichard <jguichar@cisco.com>]

Juha,

it seems to me that as the SP part of the picture is MPLS (as opposed to
ATM) then the DTE to DCE part of the VC (PE to CE) is essentially unchanged
and therefore such mechanisms as lmi and inverse-arp should continue to
function as-is. As soon as the PE 
activates a new dlci, this should be reflected within the lmi and therefore
the router (CPE) will signal using inverse-arp for the protocol address at
the other end of the VC. If this is not the case then some kind of static
frame-relay map would be needed on the CPE ..Jim

At 15:01 23/11/2000 +0200, Juha Heinanen wrote:
>Metz, E.T. writes:
>
> > When CE and PE configuration are
> > not aligned on this point, the CE may expect to connect to one site on a
> > particular circuit, while the PE actually connects it to another one.
>
>in fr or atm world, the ce learns via lmi or ilmi about new or deleted
>labels.  then it sends an inarp message along the vc and learns who is
>behind that vc and starts to exchange ospf messages with it.  the vc
>topology doesn't need to be full mesh and still all the vpn members can
>share a common ip subnet.
>
>as long as the ce-pe protocol remains fr or atm and mpls is used only in
>the provider network, there should be no changes whatsoever to the above
>described ce operation.
>
>but if the ce-pe protocol is native mpls, then the standard fr or atm
>inarp procedure stops working, because (a) mpls vcs are unidirectional
>and (b) can be merged.  one possibility might be that the destination ce
>that receives an inarp request for an unkown source, sends the reply
>back to all vcs that are still unknown to it.
>
>-- juha
> 


Jim Guichard CCIE #2069
Technical Advisory Consultant EMEA
Global Solutions Engineering 

+44 208 756 8806
Mobile: +44 7802 809763