Optimize BindBIND 9.1.0 for greater "throughput"??

Nate Campi nate at wired.com
Tue Feb 5 21:43:41 UTC 2002


On Tue, Feb 05, 2002 at 08:23:25PM +0000, phat_phoque at salmahayeksknockers.edu wrote:
> I need to increase the # of queries Bind can handle at one time.  I'm running
> BIND 9.1.0, which comes with Redhat 7.1, locally.
> 
> I see that four "named"s are running, and suspect I could maximize my
> throughput by running more...  How can I do this?
> 
> Is there another way?
> 
> (For the curious, I'm converting 1 Million+ hostnames in a database to IP,
>  but I am bottlenecking on the DNS queries.)

See http://marc.theaimsgroup.com/?l=bind-users&m=101081439312759&w=2 for
measurements I've made on the resolution capacity of BIND 8 and
dnscache. I've been able to resolve close to 1000 queries/sec - purely
resolving records *not* in it's authority, with no cache (fresh start of
the server before testing).

BIND 9, on the other hand, performs very poorly as a resolver. I've
tested it on Solaris and Linux, threaded and not threaded, sun compiler
and gcc, sparc and intel on Solaris - and BIND 9 never does well. I
usually get around 150 queries/sec for resolution in tests (sometimes 
much, much worse).

I welcome you to compile queryperf in the BIND 9 contrib directory and
see what kind of results you get. 

Oh, and I'm not getting suckered into trying to respond to the address
"phat_phoque at salmahayeksknockers.edu"
-- 
Nate Campi     Job: hostmaster at lycos.com and root at wired.com

"If you put a billion monkeys in front of a billion typewriters typing
at random, they would reproduce the entire collected works of Usenet in
about ... five minutes."   -Anon.  
 
"Come to think of it, there are already a million monkeys on a million
typewriters, and the Usenet is NOTHING like Shakespeare!"   -Blair Houghton



More information about the bind-users mailing list