Observations of an Internet Middleman (Level3) (was: RIP

Christopher Morrow morrowc.lists at gmail.com
Fri May 16 19:30:21 UTC 2014


On Fri, May 16, 2014 at 3:15 PM, Matthew Petach <mpetach at netflight.com> wrote:
>
>
>
> On Fri, May 16, 2014 at 11:52 AM, Christopher Morrow
> <morrowc.lists at gmail.com> wrote:
>>
>> On Fri, May 16, 2014 at 2:47 PM, Blake Hudson <blake at ispn.net> wrote:
>> > in the context of this discussion I think it's silly for a residential
>> > ISP
>> > to purport themselves to be a neutral carrier of traffic and expect
>> > peering
>> > ratios to be symmetric
>>
>> is 'symmetric traffic ratios' even relevant though? Peering is about
>> offsetting costs, right? it might not be important that the ratio be
>> 1:1 or 2:1... or even 10:1, if it's going to cost you 20x to get the
>> traffic over longer/transit/etc paths... or if you have to build into
>> some horrific location(s) to access the content in question.
>>
>> Harping on symmetric ratios seems very 1990... and not particularly
>> germaine to the conversation at hand.
>>
>
> Traffic asymmetry across peering connections
> was what lit the fuse on this whole powder keg,
> if I understand correctly; at the point the traffic
> went asymmetric, the refusals to augment
> capacity kicked in, and congestion became
> a problem.

Is it that? or is it that planning at some ISP pair had a '6 months to
upgrade' regularly penciled in, then 'all of a sudden' their links
were filling up faster than every 6months and... now they are 1x or 2x
upgrade cycles behind?

I imagine that up to a point upgrading a router that does only
'peering' (SFP) is 'easy', but at some step function of upgrades on
the edge ports you need to provision more backhaul and more core and
probably upgrade the link types and the chassis and ...

At some ISPs this process involves more than 1 dude/group. So
coordination and budget issues and scheduling ... become a bit harder.
Adjusting to the new reality of 'you need to plan for pipe filling
more often, increase upgrade cycle crank speed!' seems like at least
one problem, to me at least.

It's really hard to tell what's upsetting people about this whole topic :(
There's a mix of 'my access link blows' to 'isps should peer better
and for freer' and a bunch of other stuff all mixed in the middle :(

> I've seen the same thing; pretty much every
> rejection is based on ratio issues, even when
> offering to cold-potato haul the traffic to the
> home market for the users.

yes, welp... it's often rough to get folk who want to think in terms
of apples to suddenly thing in terms of the new best fruit 'acai
berry'. Especially at large and entrenched organizations.

> If the refusals hinged on any other clause
> of the peering requirements, you'd be right;
> but at the moment, that's the flag networks
> are waving around as their speedbump-du-jour.

sure, it's also super easy for them to do this, see entrenched org
comment above.
it seems to me that the point of peering is not <stalin voice>'ratio
or bust'</stalin voice> but 'mutual benefit'. If a skewed ratio of
100:1 in a local market still is cheaper than 'backhaul that traffic
from LHR to SFO' there's mutual benefit and a reason to peer.

I understand that this is a bit of a rosy landscape I'm painting, but...

> So, it may be very "1990", but unfortunately
> that seems to be the year many people in
> the industry are mentally stuck in.  :(

oh, entrenched. I see.

thanks!
-chris

>



More information about the NANOG mailing list