Fw: Peering vs SFI (was Re: Cogent/Level 3 depeering)

2005-10-06 Thread Michael . Dillon

 Time to quote Geoff Huston one more time.
 
 A true peer relationship is based on the supposition that either party 
 can terminate the interconnection relationship and that the other party 
 does not consider such an action a competitively hostile act. If one 
 party has a high reliance on the interconnection arrangement and the 
 other does not, then the most stable business outcome is that this 
 reliance is expressed in terms of a service contract with the other 
 party, and a provider/client relationship is established

Those people less versed in the art of peering may
not understand why the peer who has been disconnected
does not consider the action to be competitively hostile.

Simply put, in a true peer relationship, the party who
terminates the interconnection is shooting themselves
in the foot and inflicting as much commercial pain on
themselves as they are inflicting on their peer. The
reason that it is not competitively hostile is because
it does not increase the ability of either peer to
compete. Rather, it damages both of them equally because
true peers are equals to begin with.

As vijay points out, this whole issue is not really about
true peering because such equality between peering partners
is rare. It's really about the business case for settlement
free interconnect and that is rather more complex than 
merely the choice between free traffic exchange and
paid transit.

--Michael Dillon

P.S. would the Internet be worse off if all traffic 
exchange was paid for and there was no settlement
free interconnect at all? I.e. paid peering, paid
full transit and paid partial transit on the menu?



Re: Fw: Peering vs SFI (was Re: Cogent/Level 3 depeering)

2005-10-06 Thread JC Dill


[EMAIL PROTECTED] wrote:

P.S. would the Internet be worse off if all traffic 
exchange was paid for and there was no settlement

free interconnect at all? I.e. paid peering, paid
full transit and paid partial transit on the menu?


This assumes that one party wants to receive the bits more than the 
other party wants to send them, or visa versa.  When users on network A 
request data stored on servers on network B, which network should bear 
the brunt of the cost of this transit?  Why?  Why shouldn't each network 
(no matter their size) *equally* bear the costs for transmitting the 
data between their respective customers that their customers both want 
transmitted?


Settlements are an artifact of long distance phone service where each 
call was metered and the bill was paid by just one party.  In that 
environment it made sense for the network that was paid for the LD call 
to pay the other network a part of the fee collected for the call.  But 
the internet traffic doesn't work that way.  Users pay a fee for their 
unlimited connection to their ISP.  Content providers pay for the 
bandwidth their servers require to send the content to users.  Why not 
have each user's network bear their own costs of sending/receiving data 
between other networks.  It's data their own customers are PAYING THEM 
TO TRANSMIT.


The problem is that they each want to push some of their own costs off 
on the other party whenever possible.  Only when they are both roughly 
the same size can they not get away with bullying the other party into 
paying more than their fair share and thus we have settlement-free 
peering between the large Tier 1 networks.


In every case I've seen when peering was cut off, it was always a case 
of a big bully trying to force a smaller party (smaller network) to pay 
more than their fair share of the costs of the total transmission.  It 
doesn't matter if the smaller network is mostly content or mostly 
eyeballs or which direction most of the bits flow - all that really 
matters is that the bigger network is big enough to force the issue and 
make the smaller network pay for transit (if not paying them, paying 
*someone*) and ultimately bearing an unfair proportion of the total 
costs of transmitting the data between their two networks, between their 
two customers.


Note:  I'm all for eyeball-providers to require network providers who 
are mostly content providers to cold-potato route their bandwidth 
consuming data to the eyeball-provider's connection point closest to the 
user requesting the data, rather than hot-potato handing it off at the 
connection point closest to the content provider and leaving the eyeball 
provider with the burden of backhauling the data to their user 
requesting the data.  However, there's a HUGE difference between 
requiring cold-potato routing for content, and cutting off peering 
entirely and forcing the content provider to pay TRANSIT (to someone) to 
send the data to the eyeball provider as is apparently the current 
situation with L3 and Cogent.


jc