stress??

[Luis Daniel Lucio Quiroz <lu...@gmail.com>]

Hi all,

what does do more stress to SA,

5000 emails of 1k each one

or

1000 eamils of 5k each one

TIA

Re: stress??

[Matt Kettler <mk...@verizon.net>]

Luis Daniel Lucio Quiroz wrote:
> Yes, you are right
>
>
> Im doing my test, I have bayes and net tests, how ever it is running
> over a flash card with minimal I/O, so I guess dns will be a little
> problem then.
>
>
> Does SA has cache to remember last dns request for example?
No. It would be silly to do so. Run a caching DNS server on the same
box, and configure resolv.conf to use 127.0.0.1.

Why have SA reinvent the wheel? It would probably end up less efficient
too, since SA is coded in Perl, and most DNS servers are complied C code.

Re: stress??

[Per Jessen <pe...@computer.org>]

Luis Daniel Lucio Quiroz wrote:

> Yes, you are right
> 
> Im doing my test, I have bayes and net tests, how ever it is running
> over a flash card with minimal I/O, so I guess dns will be a little
> problem then.
> 
> Does SA has cache  to remember last dns request for example?

No, but any decent operating system will have an 'nscd' daemon for
instance.


/Per Jessen, Zürich

Re: stress??

[Luis Daniel Lucio Quiroz <lu...@gmail.com>]

Yes, you are right

Im doing my test, I have bayes and net tests, how ever it is running over a 
flash card with minimal I/O, so I guess dns will be a little problem then.

Does SA has cache  to remember last dns request for example?

TIA

On Sunday 04 January 2009 20:27:20 Matt Kettler wrote:
> Luis Daniel Lucio Quiroz wrote:
> > Hi all,
> >
> >
> > what does do more stress to SA,
> >
> >
> > 5000 emails of 1k each one
> >
> >
> > or
> >
> >
> > 1000 eamils of 5k each one
> >
> >
> > TIA
>
> Depends on what you have enabled, and how fast your network/dns and disk
> I/O are.
>
> If you have network tests enabled, the 5000 emails will put more
> pressure on your DNS than the 1000 will, just because of the RBL lookups
> in the headers. The same could also be said of all the header tests, but
> these are generally lightweight, so they make little difference.
>
> The larger messages will put a bit more pressure on the bayes subsystem
> (if enabled) and body tests. If your disk is slow and you've got bayes
> enabled, this could be worse than the extra DNS lookups, but it's unlikely.
>
>
> However, on most systems I would expect the 5000  1kbyte emails would be
> slower to process than 1000 5kbyte emails. YMMV, and you could test this
> pretty easily using time on the mass-check tool.

Re: stress??

[Matt Kettler <mk...@verizon.net>]

Luis Daniel Lucio Quiroz wrote:
> Hi all,
>
>
> what does do more stress to SA,
>
>
> 5000 emails of 1k each one
>
>
> or
>
>
> 1000 eamils of 5k each one
>
>
> TIA
Depends on what you have enabled, and how fast your network/dns and disk
I/O are.

If you have network tests enabled, the 5000 emails will put more
pressure on your DNS than the 1000 will, just because of the RBL lookups
in the headers. The same could also be said of all the header tests, but
these are generally lightweight, so they make little difference.

The larger messages will put a bit more pressure on the bayes subsystem
(if enabled) and body tests. If your disk is slow and you've got bayes
enabled, this could be worse than the extra DNS lookups, but it's unlikely.


However, on most systems I would expect the 5000  1kbyte emails would be
slower to process than 1000 5kbyte emails. YMMV, and you could test this
pretty easily using time on the mass-check tool.