Bind 8 and 9 specs.

David Botham dns at botham.net
Wed Feb 19 21:24:52 UTC 2003



> -----Original Message-----
> From: bind-users-bounce at isc.org [mailto:bind-users-bounce at isc.org] On
> Behalf Of Steven
> Sent: Wednesday, February 19, 2003 11:11 AM
> To: comp-protocols-dns-bind at isc.org
> Subject: Bind 8 and 9 specs.
> 
> Can anyone tell me what and if there is a certain number of requests
> that a Bind 8 and or 9 cache only server can handle per/sec per/min???

One thing that comes to mind here is that you have asked about a caching
only name server.  That implies that the server in question would have
to first iterate to get the answer.  Once it had the answer, it could
give it to the requesting client.

Therefore, there are two speeds to measure.  One of them, the iterative
process, has nothing to do with the speed of your server.  Delay in the
iterative process is introduced almost entirely from the servers
supplying the answers to your caching only name server.

The other speed to measure is that which with your server can answer a
query to which it already has the answer.  I believe this measurement is
comparable to the speed with which you server could supply the answer to
a query for which it is authoritative.  In other words, once the answer
is in memory, how many queries per second can your server handle.

Having said all that, I have done some testing with the queryperf
utility included with the BIND distribution.  In my lab I set up 4
systems all connected via an Extreme Blackdiamond switch (all running
100Mb/s Full Duplex).  All systems are in the same IP subnet.

The systems are:
genq1
genq2
genq3
bind1

As the names imply genq1, 2, and 3 are running the queryperf utility.
They each have a query table of 1,000,000 queries to perform.

bind1 is a bind 9.2.1 server running a PIII 1.4 GHz with 1Gig of ram.

I trigger all three of the genq boxes at about the same time and then
add up the results at the end.  I also conduct the test by running one
and then two of the genq boxes against the bind server.  In all of the
tests, the aggregate number of queries per second is about 8500.

I also did some testing with "desktop" class PII boxes and the query per
second rate dropped to about 2000q/s.

The test is fairly easy to setup and I followed the testing guidelines
in the queryperf readme...


Hope this helps,


Dave...




More information about the bind-users mailing list