DNS Query Question

Dennis Dayman dennis at thenose.net
Thu Jan 18 20:11:45 UTC 2007


David Ulevitch wrote:
> 
> Dennis Dayman wrote:
>>
>> I have a customer having some DNS issues. They have done some research
>> regarding some DNS timeout errors they saw with Verizon's sender verify
>> looking up their MX records. What they have discovered is their 
>> current DNS service has a 1% failure/timeout rate. They are exploring 
>> other vendors (UltraDNS for one), but need an estimate of the number 
>> of DNS queries for accurate pricing to put together a ROI argument for 
>> the
>> switch.
>>
>> I have no IDEA if this can be determined, but what is a good estimate of
>> the number of DNS queries generated from sending an email? 
> 
> That's not a good tack to take to figure out the answer.
> 
> Just check the logs of your current DNS server and count 'em up.
> 
> UltraDNS isn't cheap.  But neither is downtime, I suppose.

Here's what Chuq figured.

If I’m sending from my machine to your machine, here’s what I think is 
the right sequence.

HELO foo.com  (generates a call to the IP of the socket to compare to 
foo.com)

It’s also going to look up the foo.com to make sure it resolves

MAIL-FROM – it’ll look up the domain to make sure it exists, I believe.

So I think the baseline is 3, plus whatever anti-spam a site might use: 
DKIM, Sender-ID, SPF all generate at least a lookup of a TXT record, and 
depending on how they’re implemented, maybe an A. Some of the anti-spam 
stuff might pull MX to verify a return path exists, too.

I’d say the minimum is 3, max is around 8, assuming nothing cached 
anywhere, for a new connection with one email sent. Multiple emails on a 
connection helps, and pipelining helps more (but individually optimized 
emails hose that); client side caching helps a lot but we can’t depend 
on it.

If they want to send a message back (DSN, say), that’s going to pull the 
A record, then the MXes, and then for each MX, I believe it does a 
reverse lookup to get the name, and that iterates for every MX until 
sent or you run out of MXes.

-Dennis




More information about the NANOG mailing list