Re: [Idr] I-D Action: draft-ietf-idr-rs-bfd-02.txt

[Robert Raszuk <[e-mail]>]

Thank you Gert for the info !

Like Job observed for some customers peering to/via IX is just an
additional optimization hence indeed some may not always put most expensive
and latest gear there.

Actually after this discussion I am quite curious how many nets on the
routes servers really have more then one bgp path. If not much then this
entire effort here may not be practically useful.

Now if you say that the equipment is rather dated on the customer edge how
and when do you think it will get upgraded with brand new images (provided
it is not already EOE/EOL) to even support what is being discussed here ?
Shouldn't we at least try to come up with solutions which require no new
upgrades for IX customers ? At most few lines of configuration of existing
equipment ?

Cheers,
r.


On Wed, Mar 15, 2017 at 10:26 PM, Gert Doering <[e-mail]> wrote:

> Hi,
>
> On Wed, Mar 15, 2017 at 09:09:11PM +0100, Robert Raszuk wrote:
> > > - IXPs tend to get a rather entertaining mish-mash of equipment... and
> > thus bugs.
> > +
> > > Please enlighten yourself what people use "as peering router" today :-)
> >
> > While this is GROW question ... perhaps it would be very helpful to get
> > some more real data on what type of routers customers of the IX use to
> peer
> > within IX ? Why such equipment is considered at best as 2nd class
> citizen ?
>
> "What people can afford, and which can barely keep up with the demand".
>
> So, you'll still find Cisco Sup720 or Juniper MX80 as peering routers -
> reliable workhorses as far as forwarding capacity is concerned, but CPU
> starved (= keep the number of BGP sessions down if you want any
> semblance of stability), and unfortunately also with centralized BFD
> (at least on the Cisco).
>
>
> > For my BB IX peerings I would in fact put the best routers to peer
> outside
> > of my network on the edge - not just some junk or homebrew bgp/igp.
>
> If you keep buying "best of breed" boxes every 3 years, while the
> competition keeps chugging along with half the CAPEX needed, your
> customers will wander over, and you'll have to reconsider.
>
> For a "smallish" ISP over here, a "fat pipe" to an IXP is 10Gbit/s or
> maybe 20 Gbit/s - so if you need to by a US$ 50k router for that,
> every 3 years, it does quite add to the peering costs.
>
> We typically have budget to forklift IXP routers every 8-10 years...
> bandwidth growth isn't that fast anymore, so "nice and shiny" today
> will be "old and crappy" in 6 years, but as long as it keeps working,
> it will be kept around.
>
>
> It's not like "peering routers are treated second-class", more like
> "the market expects that this all costs nothing" - so *inside* the
> networks, "proper routers" get replaced by "dumb and fast" P-routers
> with little/no BGP, to save even more CAPEX.
>
> Gert Doering
>         -- NetMaster
> --
> have you enabled IPv6 on something today...?
>
> SpaceNet AG                        Vorstand: Sebastian v. Bomhard
> Joseph-Dollinger-Bogen 14          Aufsichtsratsvors.: A. Grundner-Culemann
> D-80807 Muenchen                   HRB: 136055 (AG Muenchen)
> Tel: [phone]           USt-IdNr.: DE813185279
>