scaling of DNS & mail servers

[Brad Knowles]

At 10:42 AM -0500 3/30/01, Thomas Duterme wrote:

>  Just a quick question. I have a series of small qmail servers (6 of
>  'em so far) and our organzation is scaling mail linearly for now.
>  (I don't see us getting above 6 any time soon though...)
>
>  My question: each of these mail servers is a pretty low end box
>  (PIII- 800, 256 K RAM, pizza box server).  In the past, I had them
>  all pointing to my slave server (a nicer Dell 4300 512 RAM 600 CPU
>  server).  But these boxes are now being used by marketing for
>  promotions to our user base (a series of in-house lists)

	In scaling mail servers, CPU is probably of least importance. 
Next least important is RAM.  Of most importance is the disk 
subsystem configuration, and the configuration of the filesystem on 
top of the disk subsystem.

	If you want to read about how to scale mail servers, see 
<http://www.shub-internet.org/brad/papers/sendmail-tuning/> (the 
"Sendmail Performance Tuning for Large Systems" paper I presented at 
SANE'98) and realize that while the title of the paper is about 
sendmail, probably 80-90% of the items discussed are applicable to 
all MTAs I know of.

>  So each of those mail servers runs into periods of high usage..
>  therefore highlookups.  I read earlier on another thread someone
>  recommending a very simple and clean method to scale DNS with mail.
>  Each mail server gets its own caching dns server.

	Indeed, this is the configuration recommended by Nick 
Christenson, who now works as a Senior Developer at Sendmail, Inc 
(note that Nick was my co-author on the paper "Design and 
Implementation of Highly Scalable E-mail Systems" which I presented 
at LISA 2000 in December).  However, I have some disagreements with 
this configuration, although I have agreed that I will no longer 
publicly discuss the details of my suggested alternative.

	Suffice it to say that I believe that there are some problems 
with this exact configuration.

>  I'm planning on implementing this and locking the box down properly
>  (no queries from anyone but itself), but I wanted to hear what others
>  had to think about this type of scaling.  Are there any other ideas on
>  how to scale mail and dns properly?

	I strongly recommend that you also look at the paper I presented 
at LISA 2000, by visiting 
<http://www.shub-internet.org/brad/papers/dihses/> and also read the 
"MTA Review" stuff that I was forced to leave out of the presentation 
as it was given at LISA.  Then read the papers referenced in my 
bibliography.

-- 
Brad Knowles, <brad.knowles at skynet.be>

/*        efdtt.c  Author:  Charles M. Hannum <root at ihack.net>          */
/*       Represented as 1045 digit prime number by Phil Carmody         */
/*     Prime as DNS cname chain by Roy Arends and Walter Belgers        */
/*                                                                      */
/*     Usage is:  cat title-key scrambled.vob | efdtt >clear.vob        */
/*   where title-key = "153 2 8 105 225" or other similar 5-byte key    */

dig decss.friet.org|perl -ne'if(/^x/){s/[x.]//g;print pack(H124,$_)}'