On Tue, 19 Sep 2006, Lars Roland wrote:
> On 9/19/06, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
> > I would think that a qpsmtpd system might not be able to handle as
> > much throughput as a stock qmail system, due to the amount of up front
> > processing that it does. Obviously, each qpsmtpd system is going to be
> > different, depending on what plugins admins choose to use with them.
> >
> > I currently use a dual Operton 244 system with 64 bit FC3. Dual
> > hard drives (mirrored) using a 3WARE SATA controller. qpsmtpd 0.32
> > forkserver.
>
> I scan 1.2 million emails with qpsmtpd (using clamav, spamassassin and
> kaspersky) on a single ibm x336 with 2GB ram and two 3.0ghz xeons
> (netburst based, not core 2) using Debian Sarge and a single 10K SCSI
> disk. I have found that qpsmtpd-forkserver is not the best way to go
> which is why we wrote qpsmtpd-prefork (you can grab it from
> subversion).
I didn't mention that I had 4 Gig of RAM. Since we are looking at
setting up a failover host, soon, with similar specs to the first, that
would be a good opportunity to check out the prefork code.
> The averege delay for each email is less than 10 seconds, but the load
> during peek time is much higher than what you describe (between 30-40)
> - but thats not a problem for me as long as the average scan time and
> delay stays down.
This sounds encouraging. Thanks.
