anycast roots

Joe Abley jabley at isc.org
Wed Nov 17 15:05:20 UTC 2004



On 17 Nov 2004, at 08:37, Elmar K. Bins wrote:

> in alternating fashion, but I would assume "jns1" through "jns6" are
> just the individual servers of a setup called "hgtld".

I have no idea about Verisign's scheme, but in case anybody notices 
similar distribution of queries across F root servers, it may help to 
know that:

  xxxNa.f.root-servers.org
  xxxNb.f.root-servers.org
  xxxNc.f.root-servers.org
  etc

are hosts all located at the same site "xxxN". Service is distributed 
between those hosts at that site in the manner of:

   http://www.isc.org/pubs/tn/isc-tn-2004-1.html

Service is distributed between sites in the manner of:

   http://www.isc.org/pubs/tn/isc-tn-2003-1.html

I would expect people to see high affinity to a particular site for a 
particular client, and much lower affinity between individual hosts 
within a site (but still plenty to allow DNS over TCP to function 
properly). It will be interesting to see Randy's results.

Today (2004-11-17) the only global nodes are SFO2 and PAO1; all the 
rest are local nodes (where "global" and "local" are described at 
<http://www.isc.org/pubs/tn/isc-tn-2003-1.html#anchor2>). Both global 
nodes live in the same AS.


Joe




More information about the NANOG mailing list