*Thanks for quick reply. *

*I just read the mail from Matthew
**(http://lists.dragonflybsd.org/pipermail/commits/2014-June/270300.html
<http://lists.dragonflybsd.org/pipermail/commits/2014-June/270300.html>).
But still quite blur on it! I think it is better to compare the source of
PF.*

*And I noticed that **"* state and ip fragment tables are now per-cpu.", so
in ipfw of DFly, every changes will be dispatch to context of all CPUs, so
it is also per-cpu, right ?  Can we say that the ipfw in DFly also "**able
to work in a concurrent manner on many CPUs"?*

*Regards,*
*bycn82*



On Wed, Nov 26, 2014 at 11:44 AM, Jyri Hovila [Turvamies.fi] <
[email protected]> wrote:

>  >
>
> *But what does it mean? Anyone can help to explain this? what is the
> benefit to have a PF which can concurrently manner on multiple CPUs? *To
> be a bit more specific, having several CPUs processing the traffic simply
> means your firewall / load balancer can handle bigger bandwidth without
> becoming clogged. The one-threaded nature of PF has been a very limiting
> factor with OpenBSD, the "forefather" of DragonFlyBSD.
>
> -j.
>
> --
> +358-50-5632104 (24/7)
> +358-46-8822157 (backup)
> [email protected]
>

Reply via email to