Re: [exim] High Perf server

On Fri, Jul 01, 2005 at 05:28:28PM +0200, V. T. Mueller wrote:
> Tony Finch ecrit:
> >On Fri, 1 Jul 2005, Michael Haardt wrote:
> >>Thanks goodness that cool guy at Cambridge didn't just buy big hardware
> >>for Smail, but wrote an experimental internet mailer instead. I guess
> >>he didn't think of that being too hard, either.
> >Exim was written for features, not performance.

Right. I should have said:

Thanks goodness that cool guy at Cambridge didn't just buy big software
instead of Smail, but wrote an experimental internet mailer instead.
I guess he didn't think of that being too hard, either.

But that wouldn't have brought my point through.

> Michael made a contribution by suggesting something that he thinks
> could be improved. While exim-users might not be the right place to
> discuss this, I don't really see the point in smashing it with
> replies like "go and buy expensive hardware" or "things are set in
> stone the way they are".

Tony was constructive, but he really is right on the above statement,
so shame on me. Anyway, I am sure some still had a good laugh. :-)

Back to the topic: I tried message_logs = false, but MRTG shows no
difference at all, which acknowledges what I said. It's the queue itself
causing the problem, because it enforces synchronous operations in order
to be reliable. Anything else either does not cause disk transfers or
combines many modifications into few transfers, thus not hurting really.

Unfortunately, that means combining input and msglog would not help at
all, and probably even hurt. So how about my suggestion for running
with pre-created spool files for saving directory modifications, inode
allocations and (often) block allocations?

Michael