It is not about your host performance...
...
> A very dead system. Nothing unusual happening. Minimal interrupts/second
> and context switching. But there's a high load average with 60% cpu idle
> time. The top just didn't happen to catch the cpu while waiting on disk,
> it's averaging 60%. I'm sending lots of mail through the mailing list.
> It's on cablemodem now as opposed to T1 when I had the list on the P133.
> Not sure how this would cause such a high load with low cpu useage.
The cablemodem has fairly fast channel from the network to
your machine, but SLOW link the other way. (64-200 kbps)
Part of the problem is that Qmail sends all smtp sessions in
parallel, which effectively kills your already meager outbound
bandwidth (and your machine). "In parallel" meaning:
For each recipient address on EACH message it runs individual
smtp transporter process. You have 10 messages with 20
addresses --> you have 200 smtp transporters running...
> Anyone have any ideas?
You need to have a T1 connected service provider that is willing
to do the actual distribution of the smtp messages.
Then you define message routing so that all non-local traffic
is sent to that server.
> ----------
> Tracy Reed
> http://www.ultraviolet.org
/Matti Aarnio <matti.aarnio@tele.fi>
For resource conscious mailer, look at the ZMailer:
ftp://ftp.funet.fi/pub/unix/mail/zmailer/