[85088] in North American Network Operators' Group

home help back first fref pref prev next nref lref last post

Re: TLD anycast clouds?

daemon@ATHENA.MIT.EDU (Elmar K. Bins)
Wed Oct 5 04:04:11 2005

Date: Wed, 5 Oct 2005 10:00:52 +0200
From: "Elmar K. Bins" <elmi@4ever.de>
To: nanog@nanog.org
Mail-Followup-To: "Elmar K. Bins" <elmi@4ever.de>, nanog@nanog.org
In-Reply-To: <Pine.LNX.4.62.0510050041570.15781@sokol.elan.net>
Errors-To: owner-nanog@merit.edu


william@elan.net (william(at)elan.net) wrote:

> >authoritative DNS servers (and their IP addresses) and try that list
> >on the routing registries...
> 
> Assuming that you do that, what would you be your criteria to find based
> on RR if the ip is anycasted or not?

Maybe I overestimate the openness of the people and believe everybody
would drop their documentation there...

~>whois -h whois.radb.net 194.246.96.1
route:        194.246.96.0/24
descr:        DENIC eG anycast prefix
origin:       AS31529
...

The AS object gives a lot more clue then...

aut-num:      AS31529
as-name:      DENIC-ANYCAST-AS
descr:        DENIC eG
descr:        DNS anycast AS object
...
remarks:      | DENIC eG operates the .de ccTLD registry. DENICs IPv4  |
remarks:      | DNS servers are partial unicast and partial anycast.   |
remarks:      | This object covers the IPv4 anycast setups.            |
...

Maybe using route servers helps more (coffee!) :-)


> I mean lets suppose that I run dns server at AS0 and peer at location A 
> and also at location B with same ISP with AS1. I might have one dns server 
> at location A and another different one at location B, which means its 
> anycasting. But from peering perspective it would just appear as the
> same path AS0->AS1 no matter what location it is.

That's some kind of anycasting, but from a networking point of view
it's redundant links to the same "thing" ;-)

If you are lucky, you can see different next-hops in the BGP.


> Opposite to that route to the same server might be seen from different 
> peer AS# based on location because of routing policies.

Using views from route-servers that are distant from each other might
improve the picture. The RIPE RIS project may be able to help.

Cheers,
	Elmi.

--

"Begehe nur nicht den Fehler, Meinung durch Sachverstand zu substituieren."
                          (PLemken, <bu6o7e$e6v0p$2@ID-31.news.uni-berlin.de>)

--------------------------------------------------------------[ ELMI-RIPE ]---


home help back first fref pref prev next nref lref last post