On Tue, 8 Apr 2008, Jaap Akkerhuis wrote: > > We investigated that situation on request and found some F-root > > instances were receiving very high volume of queries for invalid TLD. > > I'm wondering why those ip's aren't distributing their queries across > all the roots. > > Not all resolvers are equal. Some lock on on servers, others don't
The law of large numbers should still apply. They should lock on to servers equally often... > > 24 hour traces collected between 01-09-2007 12:00 and 01-10-2007 12:00, > > on 61 root server anycast nodes > > > > C-root, all 4 instances > > F-root, 36 out of 40 > > K-root, 15 out of 17 > > M-root, all 6 instances > > > > So the data seems to be useful (but not complete). Once we got all the > > data for DITL 2008 we could try to run the same test and look for > > trends. > > This is quite interesting. F-root is also the most anycasted of this > subset... The next highest anycasted member has the least amount of bad > TLD queries. Seems to be more variability with more highly anycasted > servers. I wonder why... > > If you look at this, > > > > >Doing an analysis per root, the numbers vary > > > > > > > >C-root 19.15% > > > >F-root 46.79% > > > >K-root 10.01% > > > >M-root 20.96% > > You'll note that C & M has less instances then K, but more of this > traffic. To me it seems too little data conclude anything. I did notice that. There is more variation in this statistic with greater anycasting than there is with less anycasting: F and K differ by 36.78, while C and M differ by 1.81. This is quite interesting and needs more investigation. But I agree there is little to go on and more data is necessary. > Do you have the number of anycast instances for all 13 roots? > > As far as I know public information is available at > http://www.root-servers.org/ There is a lot missing from root-servers.org: For example, it's not clear from the root-servers.org what roots just have backups at other sites that they turn on when the primary fails, vs what roots are anycasted. Those with 30+ sites are probably anycasted, but who can say about those with just 2 or 4 sites? Also, its never been clear what the chain of authority has been to approve anycasting. I've found some RSSAC meeting minutes, but no written approvals. The status reports to DoC are unclear as well. Who approved this practice, and on what scientific basis did they make that approval? There is also an economic question: Root server operators are not supposed to charge for root server operations. This seems to be violated by anycast operators selling anycasted root servers. Its not clear who approved that change. If you know the answers to any of these, I'd appreciate knowing. Thanks, --Dean -- Av8 Internet Prepared to pay a premium for better service? www.av8.net faster, more reliable, better service 617 344 9000 _______________________________________________ DNSOP mailing list DNSOP@ietf.org https://www.ietf.org/mailman/listinfo/dnsop
