On Wed, Apr 04, 2001 at 12:33:56PM -0700, Brett wrote:
> I just ran a test on our machine here and the results are not good. I sent a
> message bcc'ed to a 1000 different non-existent recipients on another one of
> our machines. 14 minutes later and only 600 of them have been
> processed/bounced. This is pretty slow.
You are hitting two problems here:
1) is the max number of parallel connections the remote will accept
2) you are getting only bounces back that the sending qmail has to process
which will eventually slow down the remote delivery
Better would be to configure the receiving mail server as a data sink
that will deliver (for that test!!!) messages to non existing user
to /dev/null
I have a mailing list run by ezmlm (so nearly no bounces at all) with
about 93000 subscribers on a dedicated machine. Earlier this week I did
some graphs on the delivery behaviour. It's a vanilla qmail patched with
the big-concurrency mod to get a concurrencyremote of 500.
In case you're interested the graphs are at
http://www.lamer.de/maex/creative/software/qmail/deliveries/
Maybe I'll configure qmail to a concurrencyremote of 250 and see how
the behaviour changes later this week ...
\Maex
--
SpaceNet AG | Joseph-Dollinger-Bogen 14 | Fon: +49 (89) 32356-0
Research & Development | D-80807 Muenchen | Fax: +49 (89) 32356-299
Stress is when you wake up screaming and you realize you haven't fallen
asleep yet.
