Response Times on Different Virtual Interfaces

Barry Margolin barmar at alum.mit.edu
Fri Oct 8 01:40:50 UTC 2010


In article <mailman.394.1286416987.555.bind-users at lists.isc.org>,
 Jiann-Ming Su <su_js1 at yahoo.com> wrote:

> I'm running BIND 9.6.1_P1.  The server has multiple virtual interfaces that 
> BIND 
> listens on:
> 
> listen-on { 127.0.0.1; 172.30.0.213; 192.168.43.98; };
> 
> Sometimes I can get quite a huge difference in response time depending on 
> which 
> virtual interface I query against.  For example, most of our users use 
> 172.30.0.213.  When queries against it becomes slow, I can query against the 
> 192.168.43.98 address and get reasonable response times.  Is this a 
> limitation 
> of BIND, the version of BIND I'm running, or a misconfiguration on my part? 
>  Thanks for any tips or insights.

I remember seeing this behavior years ago when I worked at an ISP.  Our 
conclusion was that the OS (whatever version of Solaris was common about 
8 years ago) had separate queues for each virtual interface.  The queue 
for the IP used by most clients could fill up, but there would still be 
plenty of room in the queue for other IPs.

-- 
Barry Margolin, barmar at alum.mit.edu
Arlington, MA
*** PLEASE don't copy me on replies, I'll read them in the group ***



More information about the bind-users mailing list