BarricadeMX experiences

[Ugo Bellavance]

Kai Schaetzl wrote:
> Richard Lynch wrote on Tue, 24 Jul 2007 18:31:11 -0400:
> 
>> I gave before and after statistics.  It was the same hardware and with 
>> our volume the input is pretty much the same each day.  The results 
>> speak for themselves.  What were overloaded servers with huge delays now 
>> run fine.
> 
> But you didn't seem to use any additional tools or measures at MTA level, 
> did you? I'm not saying that BarricadeMX is not good or maybe even 
> exceptionally good, but I'm sure that if you use a well-balanced set of 
> milters, have greylisting, use a well balanced set of RBLs and access.db 
> and then compare with *that* the comparison will be much different. 
> BarricadeMX may still be better, but surely not as much as to what you 
> compared. You cannot compare a BarricadeMX system with a more or less 
> unprotected system. For instance, only about 10 - 15% of our incoming mail 
> is spam because most of the spam is already rejected at MTA level, without 
> BarricadeMX. And viruses almost never make it on the systems, either, 
> because they are rejected on MTA level. May not be as good as BarricadeMX, 
> but good enough, especially for ressource usage.

I've seen a few sites that were running milter-link, milter-ahead, 
greylisting, zen.spamhaus.org at MTA level, and still saw a lot of 
benefit on the resource usage.  Spam catching improvement was not as 
good, but on 3 or 4 systems that were running constantly at load about 
12X # of CPUs, they are now running around 3-4X # of CPUs.