Formal study: How many points networks share
After the thread a couple of weeks ago on how many nexuses are needed for the Internet to work, the MIB (actually a woman, but that's a different thread) asked if there were any formal studies. Have any providers done a formal study of how many points they share in common with other networks? I know most providers consider their competition, if nothing else for how peering will be done. There seems to be a tremendous amount of overlap between networks. Almost every major network map looks identical. And there are some third-party consultants selling network maps. But I was wondering if anyone had a formal study from "the horses mouth."
Almost every major network map looks identical.
this is less true for the real maps than those from marketing departments and third party dealers in bumph. yes, providers are in all the same places, at least in the states. but the details of how they get there, how they do redundancy, how they handle wet paths, ... differ non-trivially. but, to your basic question, i am not aware of anyone doing non-fluff studies of where everyone inter-connects. hard to do as all the majors' inter-connections are under nda. but i imagine that we would all be willing to reveal under nda to a trusted research group such as acri, if their result would be well anonymized. but many of us would all have to send letters to each other giving notice etc. messy, but may be worth it for quality research. hey vern, pull your head out of sigcom and think about this one. randy
I would support this. -jonp On Sat, 26 Aug 2000 22:19:17 -0700, Randy Bush <randy@psg.com> wrote:
Almost every major network map looks identical.
this is less true for the real maps than those from marketing departments and third party dealers in bumph. yes, providers are in all the same places, at least in the states. but the details of how they get there, how they do redundancy, how they handle wet paths, ... differ non-trivially.
but, to your basic question, i am not aware of anyone doing non-fluff studies of where everyone inter-connects. hard to do as all the majors' inter-connections are under nda.
but i imagine that we would all be willing to reveal under nda to a trusted research group such as acri, if their result would be well anonymized. but many of us would all have to send letters to each other giving notice etc. messy, but may be worth it for quality research.
hey vern, pull your head out of sigcom and think about this one.
randy
At 22:19 08/26/2000 -0700, Randy Bush wrote:
Almost every major network map looks identical.
this is less true for the real maps than those from marketing departments and third party dealers in bumph. yes, providers are in all the same places, at least in the states. but the details of how they get there, how they do redundancy, how they handle wet paths, ... differ non-trivially.
but, to your basic question, i am not aware of anyone doing non-fluff studies of where everyone inter-connects. hard to do as all the majors' inter-connections are under nda...
Agreed in general, Randy, but I think that the interconnection points are not the only issue here. Any point where failure probabilities are not mutually independent would be a concern to me -- and I agree, I don't know of anybody doing serious work on this (probably because it is a difficult problem). For example, instances where multiple providers happen to ride the same fiber (and are thus vulnerable to the same backhoe) are interesting, especially where the fiber is not part of a ring. Food for thought, - Scott
On 26 Aug 2000, Sean Donelan wrote: :Have any providers done a formal study of how many points they share :in common with other networks? I know most providers consider their :competition, if nothing else for how peering will be done. There :seems to be a tremendous amount of overlap between networks. Almost :every major network map looks identical. And there are some third-party :consultants selling network maps. But I was wondering if anyone had :a formal study from "the horses mouth." I missed the discussion from a few weeks ago, but is this for physical network maps? If not, couldn't this be done using routing tables, RAdb information, and some geometry a la CAIDA? Aren't most accurate physical network maps borderline classified for the most part? Many network maps include a PVC they use from a Tier-1 provider as part of a 'physical' infrastructure, which would cause their network map to be conspicuously similar to that of their transit provider. CAIDA's AS connectivity maps are the closest thing to a study that I have seen. -j
participants (5)
-
batz -
Jonathon Plonka -
Randy Bush -
Scott Marcus -
Sean Donelan