cgi-bin Scripts w/Multiple Mail Servers

Vernon Schryver vjs@calcite.rhyolite.com
Thu Oct 3 20:45:38 UTC 2002


> From: P David Schaub <dschaub@dschaub.com>

> ...
> Unfortunately, I think my traffic is going to be high enough (180,000 - 200,000/day) 
> and the server separation great enough to rule out a single instance of dccm.  I 
> believe I'm going to be stuck w/either NFS (yuck) or batch pull/push (double yuck)...

200,000 DCC transactions/day is modest if you look at only DCC costs.
As evidence of that claim, notice that
http://www.dcc-servers.net/anti-spam/dcc/graphs/?resol=1month says
the 6 (until yesterday 5) DCC servers that answer to the name
dcc.dcc-servers.net have been averaging 912,000 DCC operations per day
for the last month.  One of those servers is 192.188.61.3.  It is
connected to the Internet by an puny ISDN line.  This week that server
has averaged 93,000 DCC operations/day.  Between 20:00 and midnight
last night it averaged about 200,000 operations/day.  200K operations/day
is only 2 or 3 operations/second.  Last year Luke Crawford tested dccd
and a DCC client on a single computer at well over 100 operations/second.

That said, 200,000 messages/day wants more than one dccm process
for the same reasons you probably have more than one MX server.
Today's commodity hardware can easily run sendmail at 200,000
messages/day, but you really don't want all of your mail eggs in a
single sendmail or dccm basket.  

That you wouldn't want a single dccm is especially true if the dccm
process would not be near all sendmail processes as the packets fly,
or potentially fail to fly.


Vernon Schryver    vjs@rhyolite.com

P.S. DCC clients pick the fastest DCC server.  I can't entirely explain
  why the less well connected among the 6 public DCC servers get as
  much traffic as they do.  A clue may be that about 0.5% of the
  outfits using the public servers today account for more than 50% of
  the load, which tends to make it flap among the 6 servers in chunks
  instead of more smoothly.

P.P.S. I've been given to understand that at least one competetor to the
  DCC among body filters is 100 to 1000 times slower.



More information about the DCC mailing list

Contact vjs@rhyolite.com by mail or use the form.