The question above is rather interesting, and what I'd like to know is what are the various types of "peering" that various ISPs have arranged between each other?
PEERING == Equal relation ship wher provider A announces routes to themselves and their customers to provider B and Provider B does likewise. There is no announcement by either party of non-customer 3rd party routes to the other. TRANSIT == Provider A is a customer of provider B. Provider A announces routes to itself and it's customers. Provider B announces a complete routing table and/or a default route to provider A. These are separate and distinct entities from Exchange points... Exchanges (places where traffic gets handed off) come in the following varieties: PUBLIC == An exchange point where multiple providers connect with each other across a shared media, such as ATM, FR, Ethernet, FDDI, etc. Generally most public exchanges are run by a third party to whom all of the ISPs connecting have a customer relationship. PRIVATE == A circuit or virtual circuit between two providers which does not provide a shared media type of environment with multiple providers. Generally, there is no third party customer relationship, other than possibly one or both ISPs paying a telco for a circuit. There are many different financial arrangements that can be made around private peering. One of the most common ones I've seen is to bring up multiple circuits and agree that provider A pays for one circuit while provider B pays for the other. No other money changes hands. Sometimes, provider A is willing to pay for all circuits just to get better connectivity for their customers. I hope this clarifies the terms. Owen
"public" - each ISP is at a public exchange point and traffic that is destined to/from each provider moves across whatever hardware is provided at the exchange point. This could exist at all exchange points where both parties have an adequate presence. Possible additional cost if traffic that would come from another peer is shifted here and bandwidth to the exchange needs to be increased.
"private" - A connection exists either via PTP, ATM, FR, or whatever else you can think of between the involved parties. One or both parties pay for the link. One time hardware outlay. Seems expensive. Less expensive than "buying" an DS3 or higher port from the provider. Who buys from whom? Who pays? These must be NDA'd. Can anyone cite some examples of what some providers *might* do to make it a fair trade when this is an NDA'd agreement?
"public/private hybrid" - Existing exchange point equipment and connections are used. Purchase ether/fddi cable from MFS to run between your equipment. Cheap? No telco charges, consolidate your pipes.
So what was the Exodus/GTEI agreement like? Did it fit neatly into one of these categories? Are there other more esoteric arrangements? Is GTEI basically trying to get Exodus as a "customer" in a round-about way? That certainly seems like an easy way to make more cash. "I'll talk to you, but only if you buy 'Peering Bundle #A13' from us at $100,000/mo. and you have no other options. We are big, and this is what you need to do to be a peer."
Is this an accurate assessment? Please set me on track if I'm misunderstanding the concepts at work here...
Thanks,
Charles
=-----------------= = | Charles Sprickman Internet Channel | | INCH System Administration Team (212)243-5200 | | spork@inch.com access@inch.com | = =----------------=
At 08:55 AM 8/14/98 -0700, Owen DeLong wrote:
TRANSIT == Provider A is a customer of provider B. Provider A announces routes to itself and it's customers. Provider B announces a complete routing table and/or a default route to provider A.
Pardon me if this seems pedantic, but it's late, I've had a hard week, and, well, I *am* pedantic. :) I would like to modify this definition such that if Provider B announces *any* route to Provider A which does not belong to Provider B or a customer of Provider B, then Provider B is giving Provider A transit. Please note that "default route" is a special case which is still covered under this definition because "default" includes netblocks not owned by the provider or the provider's customers. Furthermore, if Provider B propagates *any* of Provider A's routes to Provider C, and Provider C is not a customer of Provider B, then again, Provider B is giving Provider A transit. Full tables/default route are not the point, IMHO. The point is, did the packet originate or terminate outside both networks (counting "customers" as part of the networks)? If either end point is not inside one of those networks (or customers thereof), the someone has given someone else transit. Another pedantic point (which most people on this list understand): "customer" == "someone to whom you provide transit" by my definition. This follows because a provider will announce "customer" routes to a "peer".
From here on out, it starts getting circular and my head begins to hurt. :)
I'm sorry if this bored most of you, but people have answered this twice and still didn't get it exactly correct IMHO. Feel free to point out my errors if I've committed any. TTFN, patrick P.S. I really wanted to say "Provider B is providing Provider A ...", but it's already too confusing. :) ************************************************************** Patrick W. Gilmore voice: +1-650-482-2840 Director of Operations, CCIE #2983 fax: +1-650-482-2844 PRIORI NETWORKS, INC. http://www.priori.net "Tomorrow's Performance.... Today" **************************************************************
participants (2)
-
owen@DeLong.SJ.CA.US
-
Patrick W. Gilmore