On Sun, 1 Dec 1996 11:02:17 -0800 (PST) Dave Crocker <dcrocker@imc.org>
said:
>ps. I doubt that vanilla sendmail can get even artificial high numbers,
>due to its tendency to do single message per connection and even per
>process, though perhaps the recent round of enhancements has improved
>things.
Just put, say, 1000 RCPT TO: fields in your message, all to the same host
on the same ethernet. Run 100 concurrently on a PC with enough RAM to
avoid paging. That's 100,000 deliveries, now you just have to get them
out the door in 20 minutes and you'll have your 300k/hour. All you need
to do is assemble enough receiver machines so that your sending PC can be
kept 100% CPU busy (it has no paging and virtually no I/O to do). Hint:
using sendmail as a receiver is not necessarily a good idea! I don't see
any reason why you shouldn't be able to reach 100/sec or more. I've seen
a single LSMTP host input recipients at the rate of 1000/sec sustained
from a machine on the same ethernet (while running 1500 outbound
connections at the same time, all that on a processor which is totally
unimpressive nowadays). This isn't sendmail of course, but where raw CPU
cycles for protocol processing are concerned 100/sec is actually not that
much. It really means 100 RCPT TO:/250 OK sequences, and this isn't much
at all.
Actually, I seem to remember that CyberPromo touted 250k/hour, and they
are using sendmail. CyberPromo isn't the kind of company that would make
up these numbers, they knew from day one that they were going to be suing
AOL and other big names, being featured on major newspapers and all that.
The last thing you want to do in that case is discredit yourself with
figures you've just made up (as opposed to basement spam companies that
sell you "advanced turbocharged" mailers for $300 and make whatever
performance claims they feel they have to make to get your $300). As a
matter of fact, if your setup is optimized to spam only AOL accounts, it
is even possible that you might be able to reach these rates with
sendmail.
Eric
|
|
