Re: [chat] Re: [SLUG] Benefits of source distro (Gentoo) somewhat elusive :-)

[invisible ink <jdub@xxxxxxxxxxxxxx>]

scott@xxxxxxxxxxxxxxx:

> slug-bounces@xxxxxxxxxxx wrote on 05/08/2003 01:28:48 PM:
> 
> > <quote who="scott@xxxxxxxxxxxxxxx">
> > 
> > > If I am going it install a high load email server, every little speed
> > > optimisation counts.
> > 
> > This is false. Hardware and operating systems are cheap. Maintenance is
> > expensive. 
> 
> Tell that to the bosses and clients.
> They would much rather buy cheap slightly incompatible linux hardware, and 
> me spend days on it then buy decent more expensive hardware...

Cheap hardware is still cheap. :-) Better to throw appropriate solutions at
the problem... So, what are you going to get out of compiler optimisations?
Very little, especially when you're talking about a disk/network intensive
function such as mail. There's not a heck of a lot of "real" CPU work
involved ("real" in CPU terms -> maths). Big and chunky string processing,
not deep and intricate maths foo.

If you really need to optimise your mail servers, you have to focus on big
things like hardware or software. On the software side of things, you have
to think about how your chosen servers work. Perhaps you can choose an MTA
that operates in a less disk intensive manner, perhaps you can choose an
IMAP server that operates in a less memory intensive manner. You're actually
going to get decent wins by looking at the problem in this light.

Compiler optimisations don't solve architectural, design, or code issues.
They definitely don't solve administrative (non-code) implementation issues.
Do you think they're going to solve process or business operation issues?
Not likely. :-)

When you're optimising, go for the big wins first.

- ii

-- 
  Penguinillas Pack GNUzis