The MPLS-OPS Archive

[Roger Clark Williams <rogerw@nordlink.com>]

Dave, the answer is "yes".

When operating in what is called frame mode (think of it more or less as
normal routing) one label suffices. The router is saying "anyone who
wants to get to Destination X can come through me via any interface using
Label 6. I'll handle it, as I know how to get things to Destination
X". That label is sent out in all directions, so every neighbor uses
Label 6 when forwarding traffic to Destination X. I can't look it up
right now but I think this is also called liberal label retention. I
might have that part wrong, but the concept is right.

With cell-mode, when running MPLS over ATM, you can't do it that way. The
MPLS-aware ATM switch builds VCI tunnels based on MPLS labels, and for
many switches the specific interface pair is part of that VCI setup.
Therefore you can't issue one label that is good for all comers
regardless of interface. You have to issue interface-specific labels. The
switch is saying "if you want to get to Destination X, come to me
through interface ATM 1/1.201 using Label 6". Moreover, switches can
be relatively limited in how many labels they can disburse, and since
they may be disbursing relatively more because of having to do different
labels for each destination/interface pair, the switch is very
conservative in issuing labels. So, if you issue labels on a
per-interface basis, the switch can reuse that Label 6 on another
interface and mean something entirely different.

I hope that covers the ground for you.

Roger Williams





> X-MindSpring-Loop:
> rogerw@nordlink.com
> Date: Fri, 25 Jul 2003 20:16:42 +0100 (BST)
> From: Dave Stewart <dave_stewart1576@yahoo.co.uk>
> Subject: Re: Fwd: [MPLS-OPS]: Local significance of labels?
> To: Roger Clark Williams <rogerw@nordlink.com>,
>   MPLS-ops Mailing List <mpls-ops@mplsrc.com>
>
> Hello Roger,
>  
> Are labels unique per router or per interface?
>
> Roger Clark Williams <rogerw@nordlink.com>
> wrote:
> > Since labels are locally significant, they can be reused. Router A can use 
> > label 6 as a marker for Destination X, Router B can use Label 6 for 
> > Destination Y (assuming the two routers are not directly connected 
> > neighbors). Secondly, a label for a destination via Router A can be sent 
> > out by Router A to all neighbors, one label for all. Certain platforms 
> > cannot allocate any number of labels but are limited, so saving on the 
> > number of distributed labels makes sense.
> >
> > Just some reasons I can think of right now.
> >
> > Roger Williams
> >
> > >X-MindSpring-Loop: rogerw@nordlink.com
> > >Resent-Date: Fri, 25 Jul 2003 10:45:25 -0400
> > >X-Authentication-Warning: host.secure4-hosting.net: mplsrc12 set sender to 
> > >mpls-ops-request@mplsrc.com using -f
> > >X-Sender: yvy7807@mail.earthlink.net
> > >X-Mailer: QUALCOMM Windows Eudora Version 5.1
> > >Date: Fri, 25 Jul 2003 10:35:48 -0400
> > >To: mpls-ops@mplsrc.com
> > >From: dfsd 
> > >Subject: [MPLS-OPS]: Local significance of labels?
> > >Resent-From: mpls-ops@mplsrc.com
> > >X-Mailing-List: archive/latest/6170
> > >X-Loop: mpls-ops@mplsrc.com
> > >Resent-Sender: mpls-ops-request@mplsrc.com
> > >Resent-Bcc:
> > >
> > >What is the benefit of having locally significant labels versus assigning 
> > >a single label at the edge that stays with the packet as it transverses 
> > >the entire network? Why are labels swapped at each node?
> > >
> > >-------
> > >The MPLS-OPS Mailing List
> > >Subscribe/Unsubscribe: http://www.mplsrc.com/mplsops.shtml
> > >Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
> >
> > -------
> > The MPLS-OPS Mailing List
> > Subscribe/Unsubscribe: http://www.mplsrc.com/mplsops.shtml
> > Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
>
>
> Want to chat instantly with your online friends? Get the FREE Yahoo! Messenger 

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml