[dns-operations] On server selection algorithms in dns resolvers

[Casey Deccio]

On Tue, Nov 24, 2015 at 4:06 PM, Mark Andrews <marka at isc.org> wrote:

>
> In message <22100.47646.491861.300197 at tale.kendall.corp.akamai.com>,
> David C La
> wrence writes:
> > Mark, please correct me if I am wrong (as I am sure you will!) but as
> > I recall BIND9's adb didn't make a distinction between the SRTTs of
> > different addresses on a name, but rather calculated an SRTT for the
> > name as a whole.
> >
> > Thus this:
> >
> >     ns1.example.com A 192.168.1.1   (1 ns typical RTT)
> >                     A 10.0.0.1      (100 ms typical RTT)
> >
> > would behave differently from:
> >
> >     ns1.example.com A 192.168.1.1   (1 ns typical RTT)
> >
> >     ns2.example.com A 10.0.0.1      (100 ms typical RTT)
> >
> > in that the former definition would see the 10.0.0.1 address used much
> > more frequently than the latter, because both the 1 ns and 100 ms RTTs
> > would be plugged into the calculation of the SRTT of ns1.example.com
> >
> > Still true?  (Ever true?)
>
> srtt are a property of the ip address not the name and always have been.
>

See also section 7.2 of RFC 1035:

"... Note that this information should be kept on a per address basis,
rather than on a per name server basis, because the response time and
batting average of a particular server may vary considerably from address
to address...."

Cheers,
Casey
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dns-oarc.net/pipermail/dns-operations/attachments/20151124/86182ceb/attachment.html>