Stats for bind

[Kevin Darcy]

I don't know that there has ever been a comprehensive study of this or, if
there has, it's probably out of date. Most experience in this area is
anecdotal. Moreover, there are many factors involved. Is the server only going
to be serving authoritative data, or will it handle recursive
requests? A recursive server tends to use more RAM for all of the data it
caches. On the other hand, a server while only serves authoritative data tends
to use less RAM, but may need more disk I/O (to speed up reloads if it has many
thousands of zones) and/or CPU and/or network bandwidth (if processing frequent
and/or large incoming/outgoing zone transfers). From the standpoint of
performance and capacity, choice of OS shouldn't matter much between various
modern flavors of Unix (dunno about NT since I've never used it), but more
importantly you should use an OS that best allows you to perform the ancillary,
housekeeping chores. In my case, I run on Solaris and I've written a
maintenance system from scratch (recently migrating it to use Dynamic Update),
as well as a script on all of the slaves which keeps their named.conf files
automatically up to date. I also have a script which crunches up the query log
and error log data on each of my internal nameservers and mails me summaries
each day. And more. Would it have been just as easy to write all of this on BSD
or Linux? Dunno. Probably would have been much harder on NT, I know that much
(since I started writing some of this stuff in the early 90's, when NT was just
in its infancy).


- Kevin

adam wrote:

> I've been posting this question for the past couple days... Maybe the lack
> of response is because there is no subject line. I think its a fairly
> important question. So I will rephrase it. Where can I find documentation
> that shows what combination of hardware(variations of i586), operating
> system (Linux, Solaris, BSD, BIND port to NT), Zone size, number and
> frequency of secondary updates, ETC... will accommodate a specific number of
> requests in a given amount of time).
> Thanks for your time
> Adam
>
> -----Original Message-----
> From: Fathauer, Adam [mailto:AFathauer at alldata.net]
> Sent: Wednesday, January 10, 2001 5:10 AM
> To: 'Adam Palitz'
> Subject: RE:
>
> Adam,
>
>         If you get any response to this, can you forward it to me? I too am
> looking for the same data.
>
> Thanks,
>
> Adam
>
> -----Original Message-----
> From: Adam Palitz [mailto:adam at palitzlaw.com]
> Sent: Tuesday, January 09, 2001 3:58 PM
> To: bind-users at isc.org
> Subject:
>
> I'm looking for documentation that describes how bind performance differs
> depending on the platform, operating system, number of name queries in a
> given amount of time, size of the zone, etc...etc.
>
> Thank you
> Adam