On Mon, Jul 24, 2000 at 11:31:05AM -0400, Michael T. Babcock wrote:
> This is what I've asked for too -- and been given "do it yourself".
Almost certainly because:
a) It's hard to arrange a reproducable set of deliveries that
can be run on qmail and sendmail. Even a couple of hours on
the Internet can change the exact same run, eg, if
AOL changes the size of the response to MX lookups, even
an identical run will generate different traffic loads.
b) It's especially hard with email because you really want
to deliver the email to the recipient. How do you do a real
life test with real-life recipients on remote networks
without spamming them?
c) It's hard because everyone's situation differs. Should you
run a benchmark in isolation from your other network traffic
or with it? Is it legitimate to gain the benefits of, eg,
DNS caching that your web browsing might pre-load?
d) It's hard to measure. What it needs is a dedicated machine
that you can generate just the email load you want, then
take measurements off the interface (or connecting
router). Many don't have the setup/skill/motivation
to set this up.
Actually, it wouldn't be that hard, you'd need a dedicated server that
you can run qmail and sendmail on. A real life set of mail submissions
and recipient addresses and you'd smarthost qmail and sendmail to an
smtpsink. You's also use a dnscache on another machine so that you
see perfect and uncached DNS traffic. But no one seems to have posted
a test like this so until that happens, I guess it's "do it yourself".
Regards.
>
> Frank Tegtmeyer wrote:
>
> > > In his measurements that indicated that qmail used less bandwidth in
> > > real-life situations than sendmail, Dan counted the DNS traffic due to
> > > sendmail.
> >
> > And I have never seen numbers, only Dan's claims. It's hard to argue using
> > them without being backed up by numbers.
>
