[j-nsp] MPLS label stack limits on MX line cards

Nitzan Tzelniker nitzan.tzelniker at gmail.com
Mon Jun 1 16:54:31 EDT 2015

Hi Saku,

I am sure it is solvable issue my concern is that in ASIC platform (
compered to network processor ) it will not be solvable in software and we
will have to wait for a one or two hardware refresh cycles  before it will
be available.
Maybe the direction should be to do some kind of hierarchy (as the
platforms with the limitation are mostly access devices )  e.g. send the
packet with up to 3 labels to the nearest aggregation PE and it will do the
traffic engineering  (something like H-VPLS )


On Mon, Jun 1, 2015 at 11:05 PM, Saku Ytti <saku at ytti.fi> wrote:

> On (2015-06-01 21:04 +0300), Nitzan Tzelniker wrote:
> Hey,
> > The default limit is 3 but you can change it to 5
> >
> http://www.juniper.net/techpubs/en_US/junos14.2/topics/task/configuration/interfaces-mpls-maximum-labels.html
> >  I heard something about increasing this limit probably for segment
> routing
> > (but it was two years ago )
> >
> > BTW
> >
> > I didn't find a reference how small platforms that support only 3 labels
> > will be able to do traffic engineering in segment routing (ACX,QFX,ME3600
> For normal forwarding and even for FRR use cases your label depth would be
> modest. But as you say, for TE you would have pathological cases where
> stack
> will be much deeper.
> I'm not convinced this is a hard problem. I don't believe the stack depth
> limit are inherently about byte count, but more about indirection count.
> And
> if we think of TE ERO as single indirection, rather than chain of hops,
> then
> we might be able to impose multiple labels as single rewrite at one go (at
> cost of inflating next-hop count)
> If platform can do IPV6 tunneling, it can impose >40 bytes, it does not
> seem
> far fetched that you could impose 10 labels instead, provided they can be
> returned in single lookup/indirection.
> --
>   ++ytti
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/juniper-nsp

More information about the juniper-nsp mailing list