[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: IMPORTANT: your message to html-tidy



On Wed, Sep 10, 2003 at 02:41:45AM -0700, Steve Lamb wrote:
> > because you have to design a system to cope with PEAK load, not minimal or
> > even average load.  if a system can't cope with (at least!) what it is
> > going to get at peak, then it is broken.  
> 
>     Yeah, and?  Several people have pointed out that this system on your
>     hardware CAN handle the peak load with ease.  In fact it could do it
>     several times over. 

actually, the only person who bothered to do any calculations pointed out that
incoming mail is not distributed evenly throughout a 24 hour period, the
majority of it comes in during the normal business hours for your time
zone.....so you'd be running 3 hours behind for every hour of run-time.

he didn't say, but it's obvious to anyone who runs mail servers in an ISP or
similar environment, that there are also peaks in usage at 9-10am (when people
get in to work), 12.30-2pm (when people get back from lunch), about 3.30-4pm
(when kids get home from school) and about 4.30-5.30pm (when people check their
mail before going home).


> > also, you can not extrapolate linearly from a light, bursty load to heavy
> > sustained load.  that does not work.  the effects of sustained heavy load
> > are qualitatively different, not just quantitively.
> 
>     Granted.  But I was pointing out that your claims that cramming each
> message through several thousand postfix rules before handing it off to SA so
> your machine could handle it were untrue.  

most of the access maps are hashed dbs.  random access & indexed lookup, not
sequential.  only a handful of rules are PCRE (and thus serial/sequential).

you miss the point, as usual.  using the access maps in postfix *reduces* load.
running SpamAssassin on the entire header & body of a message is many orders of
magnitude more expensive than an access map or RBL lookup.  the point of doing
what i do the way that i do it is to minimise the most CPU and
resource-expensive scanning.  if i can reject spam early with a resource-cheap
rule (and thus avoid an expensive SA scan), then i'll do it.


> > when you get experience on real mail servers, you may begin to understand
> > this.
> 
>     I *HAVE* experience with real mail servers.  News flash, you don't know my
> professional background nor what jobs I've taken on freelance.  You have no
> idea whether or not I have worked on a "real mail server".  You're assuming I
> haven't because I dare argue with your holiness Sir Craig Sanders of the Duchy
> taz.net.au.  I find it laughable that you cannot conceive of someone else with
> experience NOT agreeing with you.  You're wrong.  Get over yourself. 
> Furthermore, stop lying and learn how to read.  Several people have pointed
> out that the system we're talking about fits well within the bounds of what
> you had originally said.

i'm sorry you get so offended by talking with someone who actually knows what
he's talking about.  sorry, but also basically indifferent...there's nothing i
can do about your personal problems.


craig



Reply to: