BIND performance issues

Rob Z buddy.zee at gmail.com
Fri Oct 16 15:40:45 UTC 2009


Hello,
we are running a BIND 9.6.1-P1 on a SPARC T2000 / Solaris 10 for a recursive
DNS server.
I set up a simple MRTG monitoring tool which basically runs "dig ." and
parses the output to find out a response time.
The typical response time is 1ms.
Workload on the server sometimes gradually increases, but the DNS servers
response time is not increasing proportionally.
When the number of requests hits about 5500/sec the DNS response time
abruptly goes down to 400-500ms.
The load average and processor utilization do increase proportionally to the
number of requests.
The problem starts when the load average gets to about 10 and  named binary
consumes about 55% of CPU time.
There are no network or protocol errors (except  insignificant number of
udpInErrors), no heavy disk I/O.
Just curious if anybody else observed a similar behavior.
Any ideas are greatly appreciated.
--
Rob
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/bind-users/attachments/20091016/4125f924/attachment.html>


More information about the bind-users mailing list