question about dns query distribution

[Mark Andrews]

In message <barmar-93F221.13504308022013 at news.eternal-september.org>, Barry Mar
golin writes:
> In article <mailman.1259.1360341660.11945.bind-users at lists.isc.org>,
>  "Barry S. Finkel" <bsfinkel at att.net> wrote:
> 
> > > HI Lawrence,
> > >
> > > We have recursive / caching name server for our Broadband internet
> > > services. And we have 60-40 traffic ratio. I mean 60 % queries comes
> > > on primary and 40% on secondary.
> > >
> > > Why primary does not getting 100% ?

There are lots of potential reasons.  Here are some of them:

Because sometimes it takes longer that the client waits for a response
from the authoritative servers on a cache miss.  If you are validating
you also have to take into account the additional queries required to
validate.

Because sometimes the authoritatives servers are broken so the first
server returns SERVFAIL and falls back to the second server which inturn
returns SERVFAIL.

Because queries/response are sent over UDP and UDP doesn't guarentee delivery.

Because the first server may be overloaded.

Because the nameservers are listed in different orders for different clients.

Because some clients randomise which servers they use.

Because some clients lock onto a server until it fails then lock onto the
next one that works until it fails .....

Because some clients use the fastest server.

.....

> > > Is there any way to do it ? or what is the reason behind it that both
> > > servers' having queries ?
> > >
> > > BR
> > > Ben
> > 
> > In DNS there is no concept of a "primary" and "secondary" name server.
> > All of the name servers listed in the NS records for a zone are equal,
> > and any can be used to handle a DNS query (assuming, of course, that
> > each server has the zone properly configured).  BIND will use the server
> > that has the shortest response time, but I do not know what other DNS
> > implementations do.
> 
> His question is about a caching NS, not the authoritative servers listed 
> in NS records. So the distribution is dependent on how client resolvers 
> behave, not how other BIND servers operate.
> 
> -- 
> Barry Margolin
> Arlington, MA
> _______________________________________________
> Please visit https://lists.isc.org/mailman/listinfo/bind-users to unsubscribe
>  from this list
> 
> bind-users mailing list
> bind-users at lists.isc.org
> https://lists.isc.org/mailman/listinfo/bind-users
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org