Slashdot Mirror
Level 3 and Cogent Reach Agreement on Peering
Armour Hotdog writes "Level3 and Cogent have announced an agreement on a modified peering contract that provides for settlement-free peering subject to certain unspecified conditions. This is a welcome announcement considering the disruption caused earlier when Level3 depeered Cogent. After that earlier dispute, Level3 temporarily restored peering, but announced that they would once again depeer Cogent on November 9th, unless the parties could come to an agreement."
In the new agreement, there are clauses that state that Level3 can again try to charge Cogent if their traffic amount is grossly over that of Level3's. So, while this is definitely an improvement, it doesn't rid all potential future problems.
If anything, this definitely hammers home the idea of multihoming...
Oct. 28: The modified peering arrangement allows for the continued exchange of traffic between the two companies' networks, and includes commitments from each party with respect to the characteristics and volume of traffic to be exchanged. Under the terms of the agreement, the companies have agreed to the settlement-free [i.e. no-charge -- ed.] exchange of traffic subject to specific payments if certain obligations are not met.
So what happened? It's unlikely Cogent could say "Oh yeah, we'll get 50% more retail customers so as to send traffic your way." Level 3's customers squawked and Cogent insisted they wouldn't pay? (That's Internet Mutually Assured Destruction)
Everybody's a libertarian 'till their neighbour's becomes a crack house.
Oops - I guess I could have phrased that a little more clearly. The November 9th ultimatum came when Level 3 originally restored peering back on 10/07. Today's agreement supercedes that, so the danger of another depeering (between these two ISPs) has passed unless somebody (read: Cogent) violates the terms of the new agreement.
Nope. Routing over one network to get to another is called *transit*. Cogent refused to pay for transit to get to L3. So they still had plenty of peering arangements with other networks, but none of those peering arangements allowed Cogent to reach L3. (And the same could be said in reverse.)
Do you even have any idea what it is you're discussing here? This is two companies who had a business agreement, one company abused it and got smacked by the other company.
Think this is about the US? Why don't you look into France Telecom's de-peering of Cogent awhile back.
This is not an Internet thing in that it affects the entire Internet. It is an internetworking thing in that it affects the way two ISPs exchange data.
Sit your knee-jerk, loud-mouthed, over-opinionated, under-educated ass down and shutup until you can remove your head from where you have it stuck.
Important applications should not be run on single homed networks.
DNS and routing really have very little to do with each other.
Most websites of any size whatsoever not only have multiple IP addresses assigned to the site (DNS), but also multiple links to the internet across carriers (routing). A problem in either area can cause diruption to clients, but that doesn't make them the same system.
The link you provided (minus the marketing noise) sounds like a proximity based DNS solution...also not revolutionary. Many site-to-site load balancing solutions use response time for DNS queries to determine the best IP address to return. In the event of fragmentation like happened between L3 and Cogent, if only one DNS server was avaiable on the isolated cloud (Cogent), the IP returned should be a web server on the same network. Voila, no customer disruption.
In short, yes it is possible to use DNS to work around some routing issues. It's also possible to use routing to work around some DNS issues. (example: cached DNS entries leading to a dead site, but natting the traffic to a real server and then routing the traffic accordingly). In reality though, they are separate systems. There already is a need for 'multiple routing paths', at least for any web site which wants to come close to 99.999% availability.
My UID is the product of 2 primes.
The core of the Internet is not run like you think it would be. While BGP is dynamic, when and where various prefixes (network address blocks) are advertised is tightly controlled.
When you peer with an ISP, that means you only exchange their prefixes for yours. Any other networks that may be reachable via that ISP are not advertised back to you, just like they don't send your prefixes to the rest of the Internet.
Access to other parts of the Internet via an ISP is called transit -- this is what we're all most familiar with. You give your prefixes to them and they take care of exchanging it with the rest of the Internet, and they give you the rest of the Internet's prefixes.
Lets say there are two ISPs. Lets say its Alpha (a Tier 1 ISP) and Beta (a non-Tier 1). Alpha and Beta have a settlement-free interconnect agreement -- meaning they peer with each other. Remember, that means that they can only access each other directly. To go to the rest of the Internet, it means that they have to go to other providers. Alpha uses its other SFI agreements with all other Tier 1 providers to do so, Beta may use other SFI agreements and paid-for peering or transit agreements with other providers. For so-called "tier 2" ISPs, it's often a complicated mess.
Then, lets say Alpha decides that its SFI agreement with Beta is no longer in its best interest. So, they tell Beta they're going to depeer if they don't start paying for the peering.
Beta has two options to ensure that their customer base will stay functioning. They can either come to an agreement with Alpha, or come up with a transit agreement with another ISP, either another Tier 1 ISP or Tier 2.
It gets more complicated than that, but, that's the basic jist of it.