The MPLS WG Archive

[Ken Nagami <ken.nagami@toshiba.co.jp>]

Mukul, 

VCID is used on a cell based LSR not a frame based LSR. So, in this
point, the #3 is similar to the #1. The major differnce between #1 and
#3 is that ATM-LSRs are connected through ATM switches and 
VCID procedure adds some hooks to support SVCs/PVCs through ATM switches.

Regards, 
Ken Nagami

> > >3. the VCID case. Again an overlay model where there exists CLIP or LANE
> > >except here MPLS coordinates with ATM signaling to create SVCs as needed. L3
> > >& ATM routing protocols are run.
> > 
> > That's not really true. The routing situation is actually very similar
> > to #2. The ATM switches, or most of them, are not LSRs, and just run an
> > SVC routing protocol (typically PNNI). The LSRs run IP routing,
> > but don't really participate in PNNI, except to request SVC
> > set-ups.
> > 
> > The major difference between #2 and #3 is that VCID adds some hooks to
> > support SVCs as well as PVCs. Otherwise, they're fairly similar.
> > 
> 
> I think #2 and #3 are similar only from the point
> of view of mix of ATM and MPLS signalling & routing
> protocols used for setting up the channels,
> but they differ fundamentally with respect to their models. 
> While #2 is a classical overlay model, the MPLS network
> remains unaware of the existence of
> ATM switched connections, and multiple LSPs "share" the same ATM
> virtual connection using it as a virtual link. 
> 
> The #3 is a peer model whereby the LSP is tunneled through the
> ATM network using ATM signalling. The edge LSRs in this case are
> certainly aware of existance of the ATM switched connections and
> trigger the tunnel establishment for the LSP.
> The encapsulation used for #3 is similar to #1 as specified in 
> draft-ietf-mpls-atm-04.txt, rather then that of #2 which shall
> use what ever has been established by CLIP or LANE.
> I think this makes the case #3 to be similar to #1 rather then #2.
> please correct me if I am missing something.