Re: [MPLS-OPS]: Nos of Labels in CE

On 6/25/03 5:21 AM, "Alok Dube" <[email protected]> wrote:

> Aamer,
> 
> I would like to ask the following:
> 
> 
>>>> I think you want to say same next-hop rather than interface.
>>> 
>>> No, I mean the same interface. What difference does it make to me if i
> have a
>>> prefix
>>> which has different next-hops and both the next-hops are reachable to me
> thru
>>> the same
>>> interface/sub-interface (e.g. ethernet)
>> 
>> Which next-hop would you forward the packet to, in the case of multiple
>> next-hops reachable via a multi-access interface (eg ethernet)? Keep in
> mind
>> that unlike point-point interfaces you have to do a L2 rewrite that
> includes
>> a MAC address on multi-access interfaces.
> 
> agreed,
> 
> but wherever the "label switching starts" too,you have to do an L3 lookup..
Not really. You only really need to look at the ip address (lets ignore
internal loadsharing implementations for now) if the label type is
'aggregate'. In the aggregate case the label does not convey enough
information about what the forwarding behavior should be.

> 
> take the case of a packet entring a VRF with multiple interfaces in the
> VRF.(L3VPN)
> 
> how does one determine where to Label switch the packet to without doing an
> L3 lookup?

When you allocate the label, you should know what the forwarding behavior is
going to be for the label. If you need to do a arp later on, you do it and
cache it.

> 
> atleast in the case of binding labels to P2P interfaces we have an advantage
> of avoiding an L3 lookup at egress point.
> 
> 
> 
> 
> -------
> The MPLS-OPS Mailing List
> Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> Archive: http://www.mplsrc.com/mpls-ops_archive.shtml

-- 
Aamer Akhter / [email protected]
NSITE - cisco Systems