Eric Dahnke <[EMAIL PROTECTED]> writes on 7 November 1999 at 16:40:02 -0300
>
> Could someone explain how qmail manages to be faster for average msgs. I
> can't see how it would be.
Set up a benchmarking configuration and see. Best to use real-world
traffic streams; most people's ideas of what's "normal" don't seem to
match what really goes on very well.
There are a number of issues with sorting the queue by recipient
system. To start with, you have to do a HUGE number of extra DNS
lookups to determine what the recipient systems ARE. Then you have to
sort the message list, figure out which messages are the same, and
make up your rcpt-to lists. And then you probably want to limit them
to only 20 rcpt-tos, because that's what sendmail does and other
things may not handle larger lists; this drastically limits the profit
you could gain on all this work.
In a pathological case, qmail can use a lot more network bandwidth
because of the duplication of messages going to the same system. In
practice this is rarely a serious problem. Taking into account the
*decreased* DNS traffic, it's even more rarely a problem.
Finally, network bandwidth isn't something qmail tries to optimize;
it's more concerned with minimizing delivery time. As a secondary
point, it wants to minimize queue i/o bandwidth use, since that's what
a busy qmail system generally runs out of first.
--
David Dyer-Bennet / Join the 20th century before it's too late! / [EMAIL PROTECTED]
http://dd-b.lighthunters.net/ (photos) Minicon: http://www.mnstf.org/minicon
http://www.dd-b.net/dd-b (sf) http://ouroboros.demesne.com/ Ouroboros Bookworms
