>
> At 01:56 PM Thursday 8/5/99, Jim Arnott wrote:
>
>
> >Hi,
> >
> >I'm experimenting with qmail's throughput. It approximates what
> >we want to do with it. (send mail from a mail database through qmail-inject)
> >I want to see how fast qmail will queue messages.
> >
> >I have a ~433 Mhz Alpha.
> >
> >Experiment:
> >
> > script that runs qmail-inject 1000 times
> > with a 1000 byte body
> > Qmail-send is not running
> >
> >Result:
> > takes 73 seconds (13.7/sec)
> >
> >This seems a little slow to me. The system cpu is 70% idle.
> >
> >Same experiment with "cat >> out.file < 1000byte.in" takes 4 seconds.
> >
> >
> >Is this normal ? Any ideas on how to speed it up.
>
>
> First question. Did you try the experiment:
>
> cat >/dev/null <100byte.in ?
that takes 3 seconds.
BTW thats 1000 bytes.
>
>
> Second question. Why do you think that the cat command is comparable to
> inserting a structured object into a sophisticated queue system that ensures
> reliability performance at high numbers?
Show the absolute best test scenario and to demonstrate that my system is at
least working.
> If this is a rhetorical question, then what is your expectation of the cost
> of queuing a mail message and on what basis did you make that expectation?
I'm just trying to figure out what kind of throughput to expect and justify
that against what managment expects.
IMHO it seems that there is somthing strange going on when while doing the
inject test the CPU is 70% idle. (cat test is 0% idle). There seems
to be a bottleneck somewhere.
-jim
