DNS response time

phn at icke-reklam.ipsec.nu phn at icke-reklam.ipsec.nu
Thu Dec 2 22:43:08 UTC 2004

Mike B <toastyhamster at hotmail.com> wrote:
> In message <cok6gr$2lhr$1 at sf1.isc.org>, phn at icke-reklam.ipsec.nu writes
>>afaf <afaf.e at menara.ma> wrote:
>>>I wish to know if a significant number of created zones in a DNS server 
>>>can increase the DNS response time .
>>Response time will start a significamt increase when system resources
>>are exhausted ( memory,swap, network bandwidth etc). Until then there is a 
>>small increase in responsetime, but you should not be affected.
>>The only way to know is to populate a lab and start testing.
>>>Best regards;

> This is interesting. We have a particular pair of heavily used (for us) DNS
> servers. We estimate between 80-200 queries a second are being
> handled by these, unfortunately the team that specced the hardware
> have now told us they never actually benchmarked anything, it was just
> finger in the air stuff. We've been getting nervous that BIND has been
> getting near the limits of the server, however vmstat shows nothing
> untoward, and if this *is* the definitive test for BIND performance then
> it's reassured us somewhat. Are there any docs anywhere which
> explain what resources BIND uses?

Memory and mbuf's(or equivalent) top and netstat -m will tell you most.

Rich jones(hp) has done a lot of measurment on bind, i also think he has
some benchmarking stuff, try a google search

> The reason we think BIND has been pushing the box is seemingly
> random cache corruptions that only a restart will cure, we have a few
> pairs of DNS servers on identical architecture and it's only the one
> under "heavy" load that shows this, investigations are continuing.

Ecc memory used ? If not you ought to have. 

> cheers

> --
> Mike

Peter Håkanson         
        IPSec  Sverige      ( At Gothenburg Riverside )
           Sorry about my e-mail address, but i'm trying to keep spam out,
	   remove "icke-reklam" if you feel for mailing me. Thanx.

More information about the bind-users mailing list