[dns-operations] Source data about root server anycast locations?

[Joe Abley]

On 19 Oct 2014, at 08:16, William F. Maton Sotomayor <wmaton at ottix.net> wrote:

> On Sat, 18 Oct 2014, Paul Hoffman wrote:
> 
>> Greetings again. The map and lists at <http://www.root-servers.org/index.html> have a lot of data that would be valuable to analyze, particularly how close recursives are to root nodes (even if that's just geographic/political closeness). RSSAC is just getting started with publishing reports, and anycast location and proliferation has not been covered in the reports covered so far.
>> 
>> Has anyone (maybe DNS-OARC?) kept snapshots of this type of data? Any clues would be appreciated.
> 
> For the AS112 project I have used the open resolver data to try and 'count' the number of anycast nodes by unique instances.  It would probably be one more step to infer location - But I'd need to rely on people taking the advice found in RFC6304 to state their location via the TXT record.  The end-goal of the effort is two-fold:
> 
> - how many AS112 servers there are
> - where are they and how close are they to each other.
> 
> The latter would be an educated guess based on the data.  But I think this is what you are after as well.
> 
> 	https://www.as112.net/2012-survey.html
> 
> Wasn't there a draft in DNSOP which would also make anycast nodes more easily identifiable?

We published an informational note about mechanisms available at L-Root -- see RFC 7108.

Xun Fan, John Heideman and Ramesh Govindan of USC/ISI used one of those mechanisms to do a similar external mapping exercise on L-Root following their work on <http://www.isi.edu/~xunfan/research/Fan13a.pdf <http://www.isi.edu/~xunfan/research/Fan13a.pdf>> (the paper includes results on hunting for AS112 instances). John presented the results at a DNS-OARC meeting some time ago (Dublin, IIRC).

RSSAC are getting ready to publish a document relating to coordinated measurement of traffic seen by the root server system. I can't speak for others, but personally I would like to see a similar effort standardising the geographical and topological locations of root system anycast nodes, assuming there is some useful number of root server operators who would be prepared to publish such data in machine-digestible form.


Joe
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dns-oarc.net/pipermail/dns-operations/attachments/20141020/1830b365/attachment.html>