|
North American Network Operators Group Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical Re: Peering Table Question
Randy and Martin are both correct. I often use the poker analogy. In most cases if you have a backbone that is nationwide and are at 3 geographically diverse exchange points you are in good shape with about 85% of your peers. If you approach most of the peers with the argument that you add value to their network, are competent in the administration of your network and will not create an instability in their network, and have a visible trusted Senior Network Engineer (yes a bit of the Good old boy network here) then you will get peering. The last 15% if noticable the hard part. How do you convince the "big boys" to play ball. This is where most everyone gets stuck and where a typical engineer gets in over his head. Logically engineers want to improve the network and as Martin mentioned it may not actually be in the best interest of the successful backbones to open the floodgates. The ones with the best possible luck may in fact be the web hosters. One strategy Ive long contemplated is what happens if you have such critical and demanded contact that all backbones MUST connect to diliver it to their customers (eyeballs). A mix of yahoo, etrade, fidelity, etc etc would seem to be critical. Not only would backbones need to peer to receive the information but private peering to achieve a high level of QoS would be desirable. Those that DO NOT private peer (or force the content backbone to peer at a FDDI exchange) would be at an economic disadvantage and customers flee to other backbones with more liberal peering policies. Truth is that I cannot ever see a situation where the content domains would ever allow even a momentary degradation of their services to add weight to open peering policies.... even if it is in their best interests. As for Tier1. I used to remember the longer term...Tier1 and Transit free... this got shortened to just Tier1. Some people started to attempt to change the definition from a peering description to one of geographical coverage. Tier1 is a network with national to global coverage. I still dont get this definition. David At 4:22 PM +0100 4/24/00, Martin Cooper wrote: Randy Bush <[email protected]> wrote:teir-1s don't pay for routing to anywhere. tier-2s pay for routes from tier-1s and may also pay for transit. tier-1s seem to have the majority of the customers. this may be good or bad. but it's the terminology we've been using for about seven years now. of course tier-Ns, where N is greater than 1, seem to have an interest in distorting it. big surprise.In my view the tier system is based on perceived importance, which is built on peering, and ultimately marketing. It's not that tier-1s don't have to pay for routing to anywhere, it's that they're big enough not to have to give a damn about being unable to reach a /32 on the end of a piece of string in outer Mongolia -- the piece of string will come to them if it wants reachability to their customers. Peering is a poker game. The more of it you can get, the more people will want to use your services, and the more networks will want to peer with you to reach those customers. As you go along you add bigger peers and drop the smaller ones. Lather, rinse, repeat, until your network IS the Internet because you've got everyone else's customers and they all want to peer with you to get them back. The bigger you are in terms of customers, the *less* peering you want to do. I believe UU figured that one out some time ago, evidenced by its dropping the majority of its peers to force them to buy transit to gain reachability to them. Ultimately it's all about marketing, so why should it be such a surprise that the smaller players should try to redefine the terms in a bid to gain the advantage? M. -- Thank you, David Diaz Chief Technical Officer Netrail, Inc email: [email protected], [email protected], [email protected] pager: [email protected] NOC: 404-522-1234 Fax: 404 522-2191 ----------------------------------- Build 1: 46 cities nationwide -- COMPLETE Build 2: 80 OC48s Nationwide [no typo] ++ FAILURE IS NOT AN OPTION! ++ ------------------------------------
|