[c-nsp] Downsides of combining P and PE functions into a single box

[Keegan Holley]

2011/10/19 Mark Tinka <mtinka at globaltransit.net>

> On Wednesday, October 19, 2011 04:29:50 PM Keegan Holley
> wrote:
>
> > It depends on the features.  Whatever features you need
> > on the PE are always going to be there.  Whether you
> > connect your PE to a core of P routers or connect the PE
> > routers to each other doesn't affect this in most cases.
> > Mark had a good example with the need for ingress
> > re-marking, but even that is not required in all
> > networks.  Beyond that I don't see alot of reasons to go
> > with P routers unless you have a need for route/traffic
> > aggregation which many networks do need.
>
> The issue with features is they sometimes don't work, or do
> the opposite of what you expect.
>
> Pure MPLS cores are simple, and quite feature-basic.
> However, it is understood that this may not be sufficient
> justification to spend $$ a small ISP may not have.
>
>
+1 on the $$$.   Still PE is one network P+PE is essentially two networks.
 I still don't see how this adds to complexity.  For most commonly used
features the per-hop-behavior is the same on the PE router whether the
packet came from a core P router or another PE router.  Maybe I'm missing
something, but PE routers are not going everywhere and if we're strictly
talking about complexity it's easier to manage one network instead of 2.