I just did a very quick check:

1 message delivered to a list with 1,000 members

Distribution list is pretty atypical (very low AOL, MSN, Hotmail etc counts). Its a tehcnical/professional audience.

The largest number of members in the distribution list all at the same domain is AOL with 27.

Average number of members per domain is 2.6.

MAX_RCPT_TO is set to 5.

MTA is postfix with a very close to default config (no DNS check tho)

Local cacheing nameserver via DJBDNS

Nameserver is pre-stuffed with records for distribution list (previous list activity).

Postfix spool was compleatly empty prior to run.

Machine in question is a dual PII-333 with 512Meg RAM.

Spool is on the same spindle as /var/log, and is under ResierFS

Postfix is logging through syslog to localhost.

Postfix is configured for 10 parallel deliveries to responsive MXes.

Host OS is Debian/Testing.

Total time required to hand off the message (200 spool entries) and deliver the spool down to the point of only slow MX'es left: 53 seconds (might be a bit faster, my stopwatch finger was lazy).

Total number of spool entries left after 53 seconds: 16/16 targets Total number of spool entries after two minutes: 14/14/targets Total number of spool entries after 10 minutes: 14/14/targets Translation: Every slow spool entry has only one RCPT TO and the slow MXes really are slow.

Outbound bandwith is via 100bT to a couple T3s and a T1.

Relatively meaningless numbers given the artificiality of the conditions the extremely light load, and the rather unexamined distribution pattern of the membership base, but kinda interesting none the less. Surprised me too.

-- J C Lawrence claw@kanga.nu ---------(*) http://www.kanga.nu/~claw/ The pressure to survive and rhetoric may make strange bedfellows