Re: Peering Table Question
At 01:18 AM 4/20/00 -0600, Forrest W. Christian wrote:
But that isn't the case. We live in a world where both the hosters and the accessors are paying for their access. Any settlement based in that system will be unfair as it is almost impossible to put a different value on each end of the link. Who's to say for a given AS whether inbound or outbound is more important?
It is not "inbound" or "outbound" per se, it is the amount of traffic and how far you have to carry it. Simplistic example: Network A hosts big web sites. Network B has a gazillion dial-up users. The two networks peer at MAE-East and MAE-West. The web sites are in San Jose, the dial-up users are in DC. Typical TCP flow looks like this: 1500 byte packet goes from web server to MAE-West on Network A, then transfers to Network B (because of "hot potato" routing) and comes across the country to DC destined for dialup user. Then a 64 byte ACK goes from DC to MAE-East on Network B, then transfers to Network A where it rides to San Jose. In Other Words: Network B is carrying 1500 byte packets 3000 miles, and Network A is carrying 64 byte packets 3000 miles. Sounds to me like an objective, technical reason to require one network to pay another even if they are just "peering". (Unless you are Randy, in which case one is now a "customer" of the other.) Another solution would be to have Network A agree to carry the packet cross-country, relieving Network B of the need to carry big packets long distances. However, that requires useful Network B to give Network A useful MEDs. Or maybe a bazillion other possible solutions. In summary, there is nothing wrong with settlements to help off-set unequal network costs. It is a perfectly valid business practice. Nor, IMHO, does it make one network a "customer" of the other. The two networks are just trying to share everything equally, including network costs.
- Forrest W. Christian (forrestc@imach.com) KD7EHZ
TTFN, patrick -- I Am Not An Isp - www.ianai.net ISPF, The Forum for ISPs by ISPs, <http://www.ispf.com> "Think of it as evolution in action." - Niven & Pournelle (Enable? We dunt need no stinkin' enable!!)
On Thu, 20 Apr 2000, I Am Not An Isp wrote:
In Other Words: Network B is carrying 1500 byte packets 3000 miles, and Network A is carrying 64 byte packets 3000 miles.
Ahhh now I see... Network B is actually carrying ~25x the traffic for a given flow..... Thus is costs them 30x as much for the cross-country piece, and thus Network A should in some way help out with the costs. - Forrest W. Christian (forrestc@imach.com) KD7EHZ ---------------------------------------------------------------------- iMach, Ltd., P.O. Box 5749, Helena, MT 59604 http://www.imach.com Solutions for your high-tech problems. (406)-442-6648 ----------------------------------------------------------------------
On Thu, 20 Apr 2000, Forrest W. Christian wrote:
On Thu, 20 Apr 2000, I Am Not An Isp wrote:
In Other Words: Network B is carrying 1500 byte packets 3000 miles, and Network A is carrying 64 byte packets 3000 miles.
Ahhh now I see... Network B is actually carrying ~25x the traffic for a given flow..... Thus is costs them 30x as much for the cross-country piece, and thus Network A should in some way help out with the costs.
Of course this gets really interesting when you start thinking about what kinds of implications this kind of thing can have in the long term. When Network B demands that Network A pays a settlement, Network A will obviously want to find a way out of it eventually. To do that, Network A drops it's pricing for customers that 'suck' a lot of bandwidth to balance out their traffic and avoid the settlement fees. Those customers that 'suck' are most likely the customers of Network B in the first place that Network A is now targeting. So in the end, Network A ends up creating a lot of downward price pressure on themselves!! Believe me, this is not just a theory, I've been involved with this exact situation from both the backbone and the transit purchaser side, and this is really happening. It's somewhat ironic, but it seems that the cost model for the Internet is moving towards a sender pays model. Brandon Ross 404-522-5400 VP Engineering, NetRail http://www.netrail.net AIM: BrandonNR ICQ: 2269442 Read RFC 2644! Stop Smurf attacks! Configure your router interfaces to block directed broadcasts. See http://www.quadrunner.com/~chuegen/smurf.cgi for details.
You omitted the fact that the 1500 byte packets carry a bunch of paid advertising, and thus are already revenue-producing. On 20 Apr 00, at 1:42, I Am Not An Isp wrote:
Simplistic example: Network A hosts big web sites. Network B has a gazillion dial-up users. The two networks peer at MAE-East and MAE-West. The web sites are in San Jose, the dial-up users are in DC.
Typical TCP flow looks like this: 1500 byte packet goes from web server to MAE-West on Network A, then transfers to Network B (because of "hot potato" routing) and comes across the country to DC destined for dialup user. Then a 64 byte ACK goes from DC to MAE-East on Network B, then transfers to Network A where it rides to San Jose.
In Other Words: Network B is carrying 1500 byte packets 3000 miles, and Network A is carrying 64 byte packets 3000 miles.
<snipped for brevity>
In summary, there is nothing wrong with settlements to help off-set unequal network costs. It is a perfectly valid business practice. Nor, IMHO, does it make one network a "customer" of the other. The two networks are just trying to share everything equally, including network costs.
In summary, there is nothing wrong with settlements to help off-set unequal network costs. It is a perfectly valid business practice. Nor, IMHO, does it make one network a "customer" of the other. The two networks are just trying to share everything equally, including network costs.
It depends on how you look at it. If the peer was a large colo house in one city with no national/international network, then I can see this. but..... if the company has sites around the world, willing to work with the other provider on many issues and has a large network, why should they have to pay to peer? it is just that a provider wants to make money on both ends. Lets just see what the future holds... I personally think the DOJ will walk in and clear up this peering mess once and for all. And why is this? Because the large providers keep changing the rules.... and adjust them based on who they want free peering with and who they want to pay. Christian
At 10:21 AM 4/21/00 -0500, Mark Borchers wrote:
You omitted the fact that the 1500 byte packets carry a bunch of paid advertising, and thus are already revenue-producing.
And you omitted the fact that the 64 byte ACKs carry traffic by paying customers, and thus are already revenue-producing. Does the fact that both networks are being paid by their customers mean that neither should pay the other? Or that one should always pay the other? TTFN, patrick -- I Am Not An Isp - www.ianai.net ISPF, The Forum for ISPs by ISPs, <http://www.ispf.com> "Think of it as evolution in action." - Niven & Pournelle (Enable? We dunt need no stinkin' enable!!)
On Thu, 20 Apr 2000, I Am Not An Isp wrote:
Simplistic example: Network A hosts big web sites. Network B has a gazillion dial-up users. The two networks peer at MAE-East and MAE-West. The web sites are in San Jose, the dial-up users are in DC.
Typical TCP flow looks like this: 1500 byte packet goes from web server to MAE-West on Network A, then transfers to Network B (because of "hot potato" routing) and comes across the country to DC destined for dialup user. Then a 64 byte ACK goes from DC to MAE-East on Network B, then transfers to Network A where it rides to San Jose.
In Other Words: Network B is carrying 1500 byte packets 3000 miles, and Network A is carrying 64 byte packets 3000 miles.
Sounds to me like an objective, technical reason to require one network to pay another even if they are just "peering". (Unless you are Randy, in which case one is now a "customer" of the other.)
One can also look upon it that the customer of Network A (the web site) should pay Network A for the costs involved and the customer of Network B (the dialup user) should pay Network B for their costs. A transfer of data is never initiated without two parties, one offering the data and the other requesting it. You never put data online unless you want people to look at it. Yes, I can see that if a network is very large it would not want to pay money for a private interchange, but if you are already at a shared medium you should at least offer restricted routing (for instance, you offer routes for your Washington DC customers to people at MAE-EAST). The problem is more accentuated outside of the US (in my belief). For instance, I can understand that the global players don't want to offer global routes to you if you want to peer with them at Stockholm DGIX, but this situation also means that if you buy bandwidth from a Tier 1 you get (often) lousy connectivity to local sites, sometimes routes go even transatlantic even if both sites are local but at different providers. Very inefficient. If Tier 1 providers would do more local peerings with local routes this problem would be much leviated. For instance, if I peer with UUNET in Stockholm I could get only UUNET Swedish customers from them. This problem grows as Tier 1 providers (at least over here) aims more at the end customer (which is what I am seeing). -- Mikael Abrahamsson email: swmike@swm.pp.se
On Fri, Apr 21, 2000 at 05:54:48PM +0200, Mikael Abrahamsson wrote:
The problem is more accentuated outside of the US (in my belief). For instance, I can understand that the global players don't want to offer global routes to you if you want to peer with them at Stockholm DGIX, but this situation also means that if you buy bandwidth from a Tier 1 you get (often) lousy connectivity to local sites, sometimes routes go even transatlantic even if both sites are local but at different providers. Very inefficient. If Tier 1 providers would do more local peerings with local routes this problem would be much leviated. For instance, if I peer with UUNET in Stockholm I could get only UUNET Swedish customers from them.
UUnet is especially "evil" in this respect, here in Denmark, UUnet is a relative small provider, but targets end users heavily, and thus wants to peer with the larger providers here, so their end users gets fast connectivity to domestic sites. But they only advertise their danish networks, and doesn't propagate our routes into the rest of AS702. On the other hand, in countries where UUnet is the big player, like in the UK, they refuse to peer - very inconsistent ... Jesper -- Jesper Skriver, jesper(at)skriver(dot)dk - CCIE #5456 One Unix to rule them all, One Resolver to find them, One IP to bring them all and in the zone to bind them.
UUnet is especially "evil" in this respect, here in Denmark, UUnet is a relative small provider, but targets end users heavily, and thus wants to peer with the larger providers here, so their end users gets fast connectivity to domestic sites. But they only advertise their danish networks, and doesn't propagate our routes into the rest of AS702.
On the other hand, in countries where UUnet is the big player, like in the UK, they refuse to peer - very inconsistent ...
This has been standard UUnet policy since whenever. Evil is not quite the right word - that implies too much malicious forethought. "Greedy", "whinging", "selfish" spring more to mind. Peter
participants (8)
-
Brandon Ross -
Christian Nielsen -
Forrest W. Christian -
I Am Not An Isp -
Jesper Skriver -
Mark Borchers -
Mikael Abrahamsson -
Peter Galbavy