From: "Erik E. Fair" (Internet Architect) <fair@apple.com> Perhaps the easy way out is to suggest that educating the ISPs as to what constitutes good behavior at an exchange (routing system stability and reliable packet delivery) is the responsibility of the exchange operators,
This is not just the easy way out, I believe that it is the only way out. Then, we don't have the problem of all those multiple peering agreements. Instead, just one with the exchange. Also, this eliminates the problem of each ISP trying to "filter" bogus routes. Instead, the exchange operators will handle that problem. Indeed, I think that this is/was a major impetus for "policy-based" routing in the first place (reading the old RFCs), and a clear reason we need the Routing Arbiter!
and it might even be possible to enforce some interesting policies in that regard in the route servers (e.g. if you have more than N routing or BGP peer transitions per time period, the route server will refuse to peer with you for 48 hours - think of it as the hold-down or damper from Hell).
Interesting concept!
I certainly think that to the extent that the exchange operators can measure such things as routing and peer stability, it is in everyone's interest to see the numbers (except those ISPs who are unstable). Who knows? A series good reports from exchange operators about an ISP might lead to offers of private peering arrangements outside of the exchange, to the benefit of the ISP. Similar to the way that having a good credit record seems to lead to endless offers of more credit.
I understood that the RA contract provided for these measurements, and agree that they would be a good thing! WSimpson@UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson@MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2
From: "Erik E. Fair" (Internet Architect) <fair@apple.com> Perhaps the easy way out is to suggest that educating the ISPs as to what constitutes good behavior at an exchange (routing system stability and reliable packet delivery) is the responsibility of the exchange operators,
This is not just the easy way out, I believe that it is the only way out.
Then, we don't have the problem of all those multiple peering agreements. Instead, just one with the exchange.
Also, this eliminates the problem of each ISP trying to "filter" bogus routes. Instead, the exchange operators will handle that problem.
Indeed, I think that this is/was a major impetus for "policy-based" routing in the first place (reading the old RFCs), and a clear reason we need the Routing Arbiter!
Hm, policy based routing was invented to keep nasty commerical traffic of the holy and pure NSFNet backbone. Here is the problem, NSPs don't trust the RA and the exchanges to make routing and traffic flow decisions for them. This is what you are asking them to do. Now if the RA were a commerical operation, that had a responsibitly to its customers, then you MIGHT have a different story. MAE/NAP oepraters don't have time to do routing peering stuff anyway. They are just triyng to make the thigns work.
and it might even be possible to enforce some interesting policies in that regard in the route servers (e.g. if you have more than N routing or BGP peer transitions per time period, the route server will refuse to peer with you for 48 hours - think of it as the hold-down or damper from Hell).
Interesting concept!
Exponential backoff works well, too bad the code base that does route flap damping isn't really stable enough for production use in large parts of the Internet. Best results are give, when everyone runs it. Partial deployments give poor results, as per Sean's talk a year ago. (You mean Sprint was one of the first providers to do flap dampening, I thought they were Evil, yet hear they are doing good?)
I certainly think that to the extent that the exchange operators can measure such things as routing and peer stability, it is in everyone's interest to see the numbers (except those ISPs who are unstable). Who knows? A series good reports from exchange operators about an ISP might lead to offers of private peering arrangements outside of the exchange, to the benefit of the ISP. Similar to the way that having a good credit record seems to lead to endless offers of more credit.
I understood that the RA contract provided for these measurements, and agree that they would be a good thing!
As far as I know, no one has said these stats would be a bad thing. -- Jeremy Porter, Freeside Communications, Inc. jerry@fc.net PO BOX 80315 Austin, Tx 78708 | 1-800-968-8750 | 512-339-6094 http://www.fc.net
Exponential backoff works well, too bad the code base that does route flap damping isn't really stable enough for production use in large parts of the Internet.
I think the code base that does route flap damp(en)ing is in use in large parts of the Internet...
Jeremy Porter, Freeside Communications, Inc. jerry@fc.net
Avi
participants (3)
-
Avi Freedman -
Jeremy Porter -
William Allen Simpson