Amedeo Rinaldo:
> Il 08/04/2011 16:06, Wietse Venema ha scritto:
> >> .. [cut] ..
> > postscreen changes the calculation of "cost".
> >.. [cut] ..
>
>
> Really intresting point of view, i need to spend more time on it.
>
> About resource consuming .. i have to check/match my resource/snmp
> monitoring to better evaluate. I'm now using few (and quite light
> resource consuming) pcre rules and they kill about 60-80% of potential
> dnsbl-ed senders. I've given a rapid sight at system graphics now and
> during the 2 days of my postscreen dnsbl tests i've noticed more dns
> look-up and cpus resources pretty unchanged ..
Postfix uses little CPU, so that is not necessarily a good metric.
A better base for comparisons is "latency", the time to complete
operations including (especially) network read and writes.
Smtpd processes work on one thing at a time, which maximizes latency.
Postscreen works on things in parallel, which reduces latency. This
is possible because postscreen does only simple things.
> But consider the system flow has not been altered to well integrate
> postscreen (only a rapid test); so i'm sure you are right when you say
> "postscreen changes the calculation of costs" !
>
> Have someone already done fine check resource consumption comparisons?
> I'm going to play more with ..
>
> Ciao e buon week-end!
Enjouy the weekend.
Wietse
