Respectable enough, but I'm not sure why you bother having that big of
an array with that small of a mail load. I've got single machines
handling loads several times larger, all doing Clamd, a commercial
scanner, SA and more on milter during the connection time, and there are
no SMTP timeouts in the logs that are due to load on the machine. (2-3
timeouts in the logs per hour, usually due to icmp unreachables from our
replies to them)
Richard Frovarp wrote:
> Large array of mail servers? What does that mean exactly? We have a
small number of machines that scan mail and pass it on to a slightly
larger number of mail storage machines. However, each of our scanning
machines handles 150,000 to 200,000 messages per day, with our
legitimate traffic occurring between 8:00 and 15:00 local, which is
probably typical of most places. We use MailScanner with average scan
times of 7 seconds per message. Even at that, when we are hit hard (ten
of thousands of users is a pretty big target) we can have a large number
of messages waiting to be scanned. Since we aren't holding SMTP
connections open, we can keep accepting mail. I suppose if there were no
available SMTP processes to accept, most servers would attempt a
redeliver, so no harm no foul.
Our 200,000/server is no where near the 500,000 reported on the list
earlier, but it is a respectable number. Oh, and automatic ALL_TRUSTED
works for us.
--
Jo Rhett
Network/Software Engineer
Net Consonance
