Antwort: Re: [exim] Exim4 and SA: Overloading the system

[sr...@abit.de]

Patrick von der Hagen <pa...@wudika.de> schrieb am 17.01.2006 20:47:52:

> PS: more RAM usually is a good idea but situations have been reported, 
> where adding memory just killed performance, so be careful with such 
> generalizations. There have been Intel-mainboard-chipsets with 
> 2nd-level-caches supporting up to 256MB. Adding more memory gave you 
> more memory, sure. But disabled your 2nd-level-cache.....

Uhm, I wonder were you got this information from...
But advising someone to NOT upgrade his memory over 256MB
on a serversystem that certainly needs MUCH more ram is
really strange in my opinion. Exim + SA are bound to kill
the machine in anything bigger then a homeserver-environment.

I'd say 1GB are absolutely minimum for such a server.
Usually you'll be runnnig 10 spamd processes (plus the spawning
father), which take up around 50MB memory per process. Add
exim (not much) and apache up to that and you'll easily end
up with an average memory consumption of around 700MB.
That leaves you with around 300MB spare memory for the bad
days ;)

my .02€
Sascha

PS: Crossposting is bad, but doing a reply-to-all on a crosspost
is even worse.

PPS: I'll keep it on the SA list, as SA is most likely the
reason for the slowness of the machine

--------------------------------------------------
Sascha Runschke
Netzwerk Administration
IT-Services

ABIT AG
Robert-Bosch-Str. 1
40668 Meerbusch

Tel.:+49 (0) 2150.9153.226
Mobil:+49 (0) 173.5419665
mailto:SRunschke@abit.de

http://www.abit.net
http://www.abit-epos.net
---------------------------------
Sicherheitshinweis zur E-Mail Kommunikation /
  Security note regarding email communication:
http://www.abit.net/sicherheitshinweis.html

Re: Antwort: Re: [exim] Exim4 and SA: Overloading the system

[Patrick von der Hagen <pa...@wudika.de>]

srunschke@abit.de wrote:
[...]
> Uhm, I wonder were you got this information from...
It was quite some time ago... early Pentium 1, IIRC. I just hate 
oversimplifications and "adding more RAM will certainly help" is 
certainly wrong. I did mention several scenarios which might cause his 
problems without being related to RAM in the first place.


> But advising someone to NOT upgrade his memory over 256MB
> on a serversystem that certainly needs MUCH more ram is
> really strange in my opinion. Exim + SA are bound to kill
> the machine in anything bigger then a homeserver-environment.
I didn't say "don't upgrade your memory". I said "analyze the problem 
before investing money in memory". In my opinion advising someone to 
invest money in a solution that might not be related to the problem at 
all is the strange thing....
Especially if someone says "my server is P4 1.4GHz, 256 RAM and 40GB HD 
and it is NEW". IMHO that's low-budget and quite a homeserver-setup.
And I doubt that his server "certainly" needs more memory. One of my 
servers runs more software and still requires less memory than 256 MB 
RAM. It's just 2000-2500 mails a day, but I could certainly handle much 
more. His 216 users might cause more than 2500 mails/day, but it might 
as well be less than 2000.


> I'd say 1GB are absolutely minimum for such a server.
Without more information about the system-load I consider this to be a 
wild guess. Very wild.


> Usually you'll be runnnig 10 spamd processes (plus the spawning
> father), which take up around 50MB memory per process. Add
So, if I ran 10 spamd processes in parallel as an average and were using 
Spamassassin 3.1.0 which as default spawns as many processes as 
currently required and if each mail took 10s to be processed... than I'd 
process 1 mails/s or 86400 mails/day. If I had 250 users that would be 
345 mails for each user....
Doesn't sound realistic to me...
Using 10 spamd does not sound right if you have a mere 250 users. I'd 
guess less than five...


> PS: Crossposting is bad, but doing a reply-to-all on a crosspost
> is even worse.
I agree, but since there is no indication wheter Exim or SA (or 
something else) is the cause, I saw no way to choose an appropriate list.

> 
> PPS: I'll keep it on the SA list, as SA is most likely the
> reason for the slowness of the machine
I'd guess that he's scanning mail for invalid recipients or perhaps even 
created an open relay. More information might prove me wrong and prove 
you right.

-- 
CU,
    Patrick.