6.x, 4.x ipfw/dummynet pf/altq - network performance issues

[Mike Tancsa]

At 08:03 AM 2/6/2007, Chris wrote:
>On 06/02/07, Justin Robertson <justin at sk1llz.net> wrote:
>>
>>I've actually already done everything you've suggested with little or no
>>impact at all. One point where we have different results is with
>>ADAPTIVE_GIANT, I actually noticed a drop of about 50kpps thruput when
>>disabling it.
>>
>Hmm I am surprised not more attention has been brought by your
>observations, I have noticed myself freebsd 4.x holds up much better
>during ddos then 5.x and 6.x probably due its better robustness and
>lighter code.

When I ran through the tests with pmc compiled in there wasnt any 
obvious areas where it was spending a lot of time.  What I was told 
was that the locking overhead was a big penalty and more emphasis was 
put on correctness than speed going from RELENG_4 to RELENG_5 and 
above.  Supposedly the payoff will come as more CPU cores get added 
as there is better potential to scale with this design.  While I did 
see some improvement in the box with SMP compiled in, it still has a 
ways to go for this application


>I do hope but will probably be dissapointed stability and robustness
>is on the todo list for the devs in aiming to get 6.x to where 4.x
>was.

I have found stability to be quite good.  But certainly as a firewall 
or router, the speed is not there yet.

         ---Mike