[ previous ] [ next ] [ threads ]
 
 From:  sai <sonicsai at gmail dot com>
 To:  "Alex Neuman van der Hans" <alex at nkpanama dot com>
 Cc:  m0n0wall at lists dot m0n0 dot ch
 Subject:  Re: [m0n0wall] Probably completely OT but here goes (was Re: [m0n0wall] m0n0wall throughput)
 Date:  Thu, 23 Feb 2006 13:12:56 +0500
On 2/23/06, Alex Neuman van der Hans <alex at nkpanama dot com> wrote:
>
> > If you said the reverse was true, I could almost believe that (though
> > that would leave pfsense much slower than it should be).  You have
> > some serious problem if pfsense performs one iota better than
> > m0n0wall, much less 10 times faster.  FreeBSD 4.x is at least twice as
> > fast (networking-wise, on a single proc machine) as 6.x, hence
> > m0n0wall should be at least twice as fast as pfsense.  There may be
> > some limited circumstances where improved driver support reduces that
> > difference, but it would be pretty safe to say that pfsense will
> > virtually never be faster than current m0n0wall versions.
> >
> If I may ask, just for informative purposes, why is 4.x faster than 6.x? And how does multiproc
help/hinder?
>
> --
>
> Alex Neuman van der Hans
> N&K Technology Consultants
> Tel. +507 214-9002 - http://nkpanama.com/
>
Seems the FreeBSD people thought that SMP was mort important (and
current multicore cpus seem to prove them correct) and so SMP was
focus of development after 4.x
This reduced performance of the network stack.
Andre Oppermann et al are now working on the network stack to improve
performance.

sai