DNS best practice - server placement

Kevin Darcy kcd at chrysler.com
Fri Jul 25 01:22:31 UTC 2008


Barry Margolin wrote:
> In article <g6af6u$2qvn$1 at sf1.isc.org>,
>  paleale at sonic.net (Alan Strassberg) wrote:
>
>   
>> 	What's preferrable - 
>>
>> 	* An internal DNS server with forwarder statements to an upstream (ISP)
>> 	DNS for Internet resolution, or 
>>
>> 	* An internal DNS server forwarding to a DMZ DNS server that does
>> 	the upstream query. 
>> 	
>> 	Other than performance issues, it the internal + DMZ design "better" 
>> 	or is this just adding latency and more points of failure?
>>
>> 	What's the best practice architecture for a large corporation?
>>
>> 					alan
>>     
>
> Is "None of the above" an acceptable answer?  What's wrong with
>
> * An internal DNS server that does normal iterative resolution from the 
> roots.
>   
 From strictly a DNS architecture/operational standpoint, that's best, 
but it's probably the *least* acceptable to the security/auditing 
department of a large corporation. It means opening holes directly 
between the internal network and Internet. The whole point of having a 
DMZ is to avoid doing that.

I think a better question is: why is an *internal* server resolving 
Internet names in the first place, either directly or indirectly? I 
think most folks these days are going to a proxy model where the only 
things that need resolution of the Internet names are the proxies. In 
which case they can ask a DMZ nameserver, they don't need to talk to an 
"internal" nameserver to get that resolution.

                                                                         
         - Kevin



More information about the bind-users mailing list