[132715] in North American Network Operators' Group
Re: Ratios & peering [was: Level 3 Communications Issues Statement
daemon@ATHENA.MIT.EDU (Leo Bicknell)
Tue Nov 30 10:33:46 2010
Date: Tue, 30 Nov 2010 07:33:36 -0800
From: Leo Bicknell <bicknell@ufp.org>
To: NANOG list <nanog@nanog.org>
Mail-Followup-To: NANOG list <nanog@nanog.org>
In-Reply-To: <7BC3C7E5-6E99-4698-8D54-2A74E8989899@ianai.net>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
--7iMSBzlTiPOCCT2k
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
In a message written on Mon, Nov 29, 2010 at 11:47:10PM -0500, Patrick W. G=
ilmore wrote:
> Ratios were an excuse used by GTEi to try and force Exodus, Above.Net, an=
d Global Center to pay for peering back in 1998. It had a valid, technical=
reason behind it - the cost of bit-miles.[*] Unfortunately, most people h=
ave forgotten this and simply claim one side is more 'valuable' than the ot=
her. In reality, the "value" of a relationship is NOT related to the numbe=
r of bits flowing in either direction.
[snip]
> [*] 10 second explanation for those who do not understand: I hand you a s=
mall HTTP GET request, you carry it across the country. You had me a 1500 =
byte web page, I carry it across the country. My costs are much higher tha=
n yours, you need to compensate me for the additional costs.
I don't know how much GTEi specifically played into this, but this
is all one of the reasons AboveNet actually asked peers for MEDs
and honored them. AboveNet had many peers that sent useful MEDs
(typically from IGP cost on the other side) and routed based on
them. That completely flips the cost.
I agree it's important to look at such issues when peering, because
it should be "fair" for some approximation of fair. However if
folks really wanted fair they would look at technical solutions
like MEDs, selective routing, peering with regional ASN's, etc.
I'll also point out that it's my feeling that over time the issue
you describe has become less important. When we were paying $50,000
a month for an OC-3 across country the bit-mile cost was huge, and
moving those extra bits was a huge deal. Now you can get a cross
country 10GE wave for $5,000 a month.
A large part of this discussion is about the cost of providing
bandwidth at the edge. I would venture for most residential end-user
providers somewhere between 75-90% of the infrastructure cost is
from the customer prem to the "POP" in the nearest major city. Sort
of the extended last mile, if you will. The last 10-25% is the
"backbone" cost, city to city transport, peering, etc. I think
that ratio is increasing over time, in another 10 years I expect
it will be 90-95% local cost, and 5-10% backbone cost.
I said before, ratio is an outdated concept, and getting more so
by the day. That doesn't mean ignore it, or don't understand it,
but folks who are depeering based on ratio are either living in the
past, or using it as a straight up excuse for their real motivations.
--=20
Leo Bicknell - bicknell@ufp.org - CCIE 3440
PGP keys at http://www.ufp.org/~bicknell/
--7iMSBzlTiPOCCT2k
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.13 (FreeBSD)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=piH3
-----END PGP SIGNATURE-----
--7iMSBzlTiPOCCT2k--
