Determining ownership of Internet routing problems
I am a little hesitant to post this here, as it comes from the perspective of a user (albeit not a lay user). However, I believe the issue is very much one for service providers. Although it has been previously discussed on comp.protocols.tcp-ip, I have yet to determine whether there is a clear answer, let alone what it is. If I am posting inappropriately, feel free to tell me so. -------------- I had an... "interesting" problem a little while ago. I couldn't reach my mail server, and I couldn't tell who was responsible. The problem appeared to be a routing loop somewhere between my connectivity ISP and my hosting ISP. I talked to the connectivity ISP, and they said the router was outside of their network and run by someone they had no contract with. The hosting ISP said essentially the same thing. Now, I realize that dynamic routing means that there's no real way to predict the path a given packet will take. But I had somehow thought that the contractual arrangements between ISPs and their backbone providers would mean that there must be service agreements between everyone on the path between two points, and that if a link failed, there was a path of contractual responsibility. E.g. [backbone provider] / \ [intermediate A] [intermediate B] / \ [ISP A] [ISP B] where (say) ISP A is the connectivity provider, and ISP B is the hosting provider. So if I can't reach ISP B, either ISP A or B should be able to talk to his upstream provider and get it fixed. Now I'm wondering if that is even a valid assumption. Maybe the truth is more like this: [backbone provider A] [backbone provider B] / \ / \ [intermediate A] [intermediate C] [intermediate B] / \ [ISP A] [ISP B] and if the problems is with intermediate C, I'm probably SOL. Clearly, I would want my ISP to insist that his upstream providers not allow such unreliable topologies to be used. So, my questions are, am I asking too much? Am I misunderstanding the real world of the Internet? And am I posting in the wrong forum? /kenw Ken Wallewein CDP,CNE,MCSE,CCA,CCNA K&M Systems Integration Phone (403)274-7848 Fax (403)275-4535 kenw@kmsi.net www.kmsi.net
Now I'm wondering if that is even a valid assumption. Maybe the truth is more like this:
[backbone provider A] [backbone provider B] / \ / \ [intermediate A] [intermediate C] [intermediate B] / \ [ISP A] [ISP B]
and if the problems is with intermediate C, I'm probably SOL. Clearly, I would want my ISP to insist that his upstream providers not allow such unreliable topologies to be used.
Ken Wallewein CDP,CNE,MCSE,CCA,CCNA
This is a close approximation of the actual topography - though there may be a few more 'C's between A and B. The reality is that there are a couple of really large providers and the majority of networks connect either directly or are no more than a couple of hops off of the largest networks. See http://www.caida.org/analysis/topology/as_core_network/AS_Network.xml. End users trying to fix routing loops usually have fairly limited success due to the difficulty in notifying the affected providers - this is an issue that has been hashed to death and I don't think anyone is in the mood to discuss it again. Your 2 upstream providers are the ones you need to hold responsible - it's their job to get your packets where they need to go. It is entirely possible that various networks on the Internet will not be able to talk to other networks due to policy and peering decisions made by the various providers. Providers that do not provide access to the majority of the Internet usually find they loose market share quickly - this tends to be self correcting in fairly short order. Mark Radabaugh Amplex (419) 720-3635
participants (2)
-
kenw＠kmsi.net -
Mark Radabaugh