On Tue, 1 Jul 2003, Aliaga, Luis Carlos - (Per) wrote:

> Hello List,
> Does anyone has information about peering policy? Specially, in case  either
> concerned party does want to charge the other for traffic exchanged in
> either direction. We want to setup a bilateral peering with other ISP here
> in Perú.

To massively overgeneralize, there are essentially two approaches to take
with peering policies.  One is to peer with everybody you can, and the
other is to be restrictive, essentially making sure you're getting
something out of every peering session.  There are of course many networks
with policies somewhere between the two extremes.

The first approach tends to be favored by smaller ISPs.  If you're paying
for transit, any amount of traffic you can send over a peering link
instead of a paid transit link reduces your transit bill, and that's seen
as a good thing.  Of course, to justify peering as a cost savings you need
to make sure you're not paying more to get to the exchange than you're
saving in transit costs, but even in that case once you're at an exchange
peering generally makes sense.  Even in that case, some amount of traffic
analysis is generally a good idea, so you'll know which of the more
difficult potential peers are worth expending effort on.

The closed policy approach is generally seen with bigger networks, who are
already passing most or all of their traffic through peering.  If you're
already able to hand off traffic for free, bringing up new peering
sessions doesn't gain you a whole lot, and requires some effort (and thus
cost) to turn up and administer.  Therefore, for the biggest transit free
networks, turning up new peers really isn't seen as worth doing at all.
Those somewhere in the middle sizewise are also often selective -- while
they may be paying for some transit, they gain very little by peering with
somebody they're already getting to through peering.  Furthermore, if the
upstream provider of a potential peer is already a peer, and the upstream
has a requirement for how much traffic needs to be passed across the
peering session to maintain peering, peering with the peer's customer
could drop the traffic levels below the requirement.

Some networks also require the ability to peer in multiple locations,
either to avoid one network carrying traffic for long distances while the
other barely touches it, to ensure redundancy, or just as a threshold to
make sure the other network is at least some arbitrary size.

What this really comes down to is that free peering is generally not done
out of altruism, but rather out of a perceived mutual benefit.  The idea
is that since both parties are benefiting, neither of them should be
charging the other.  If one party can't be convinced that there's any
benefit to peering with the other party, they probably won't do so for
free.

There's also sometimes a game of chicken aspect to these negotiations.  If
somebody you want to peer with is trying to charge you, it may really be
that they don't see any benefit to peering and would only connect to you
if paid to do so, or it may be that they want to peer with you but want to
see if they can get some money out of you at the same time.  In that case,
you've essentially got three choices -- decide it's not worth peering with
them at that price and walk away, decide that you would pay that price if
you had to, but that they'll probably peer with you anyway if you refuse,
or decide that they really aren't likely to peer with you for free, but
that the cost is worth it, and pay.

Does this come anywhere close to answering your question?

-Steve

--
Steve Gibbard
Steve Gibbard Consulting -- IP network consulting services
Office: 510 528-1263
Cell: 415 717-7842
http://www.gibcons.com/

------------------------ANNOUNCEMENT---------------------------------
     *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-
                         ** 802.11 Planet **
       The business of hot spots. Wireless ISP business plans.
      Wireless network certifications. High-gain antenna design.
                       Site survey techniques.
                       June 25 - 27, Boston, MA
             http://www.jupiterevents.com//80211/spring03
     *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-

Replies

[isp-bgp] PEERING POLICIES, Aliaga, Luis Carlos - (Per)