At 10:28 AM 6/29/2002 +0100, Stephen J. Wilcox wrote:

...

OTOH, some networks who peered with anyone and everyone did not survive. While some networks who peered with no one have also died. (And some who peer with no one just over-report EBITDA by more than the GNP of many countries. :-) So I am not sure there is any strong evidence that peering or not is good for long term economic viability.

I doubt peering for a large "tier 1" is directly affecting their economic state, they will see all networks via peers with their fellow "tier 1" networks and peering further downstream isnt going to alter cost..

Of course they alter costs. NAP fees, OC lines, router interfaces, management of BGP sessions, etc., etc. all cost money. Furthermore, peering at 15 places probably requires less of a national backbone than peering at 4 places because you have fewer and closer exit points, which may reduce internal costs. Then again, the fees for 15 NAPs, plus 15 OC lines, plus 15 router cards, plus whatever else may well be more than the cost of increasing your national backbone. And as for being able "see all networks" just because you are a "Tier 1", do you mean like Above.Net and PSI? Or PSI and C&W when they recently had a bit of an argument? Or Exodus and BBN a while ago? Or any of a gazillion other examples? As shown on this list, just because you think you are a "Tier 1" these days does not guarantee you a view to all other networks. And it *certainly* does not guarantee you anywhere near an optimal path to all prefixes in the global Peering does affect both stability and economics. I just think it affects them differently than most "tier 1" players - the more you peer, the better your economics. They think that the less you peer, the better your economics because then people will pay you to get to your backbone.

...

I do believe there is operational evidence that a more open peering policy can reduce latency to off-net locations, but I am sure there are other reasons to close your peering policy.

I think this is the key point. Its common sense that peering with the downstreams will improve user quality of service by both reducing latency and taking unnecessary points of failure out of the network.

This is why I think peering more will help you economically. Peering less might get you a few more downstreams, but we how many people are really ready to peer at the big peering points? 200? 500? And how many downstreams does someone like UU or Sprint have? 20,000? 50,000? Probably more. KPNQwest claimed they had over 100,000 business customers before they went belly up, and I bet US carriers have more. So they are trading reduced performance and reliability, plus pissing off the rest of the Internet community, for the *potential* to sign less than 1% of their current base. (Remember, just because SmallISP.com cannot peer with UU, C&W, Sprint, AT&T, Qwest, L3, etc. does not mean he will buy from all of them.) All that said, there are other reasons for big networks not to peer with anyone who asks. Both economic and technical reasons. I just personally feel the bar is too high for most big players. Then again, I repeat, the Internet is still (mostly) regulation free, so if you own a network, you get to peer - or not - with whomever you please. And I would not have it any other way.

Steve

-- TTFN, patrick