[j-nsp] Limit on interfaces in bundle

[Jesper Skriver]

On Tue, Nov 03, 2015 at 01:22:49AM +0200, Saku Ytti wrote:
> On 2 November 2015 at 19:14, Jesper Skriver <jesper at skriver.dk> wrote:
> > Right, on those types of platforms it can be done - assuming there
> > are spare bits in the meta data that goes with the packet, enough
> > free instruction space etc - but it will come at a performance impact
> > as it will require more cycles per packet, unless on the
> > particular platform there is still headroom for adding more cycles
> > without affecting the ability to be linerate ...
> >
> > Which is why I in my original reply said that it would not be
> > practical. For it to be useful all routers in the path needs to
> > support it, otherwise as soon as it hits one that doesn't all the
> > various MPLS sub-types will get mixed together again.
> > Features which require network wide support are quite hard to get
> > off the ground.
> 
> I'm not sure it would change anything from status quo, except make the
> duck type explicit type.
> 
> Devices which today do MPLS payload duck typing, will have this
> classification information on egress NPU, as they are doing ECMP
> decision. Same network today can have other LSR devices in the path,
> which don't do duck typing, and will not ECMP the flow. This does not
> cause problems, assuming there isn't capacity problem.
> So I don't immediately see, how come this isn't entirely local decision.

The issue is not who does what with regard to ECMP hashing.

I magine you have 4 routers in a line A-B-C-D, router A sends
packets with different ethertypes depending on payload, but router
B doesn't have the ability to preserve this, so all MPLS packets
leaving B has the same ethertype. Now regardsless of the
capabilities of C and D they will see packets with the standard
MPLS ethertype and no benefit can be had.

Entropy labels is much more useful, they transparently traverse
parts of the network that has no support for them, and supporting
them doesn't require all the complexity of the proposed solution.
And not least, it has been standardized a long time ago, and many
platforms support it today.

/Jesper