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

[Justin Robertson]

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.


Mike Tancsa wrote:
> On Mon, 05 Feb 2007 14:03:41 -0800, in sentex.lists.freebsd.questions
> you wrote:
>
>   
>> I suppose my concerns are two-fold. Why is 6.x collapsing under traffic 
>> that 4.11 could easily block and run merrily along with, and is there a 
>> queueing mechanism in place that doesn't tie up the box so much on 
>> inbound flows that it ignores all other relevant traffic?
>>
>> (as a note, all tests were done with device polling enabled. Without it 
>> systems fall over pretty quickly. I also tried tests using 3com cards 
>> and had the same results)
>>     
>
>
> On the 6.x box, try enabling adding to /etc/sysctl.conf
>
> kern.polling.enable=1
> net.inet.ip.fastforwarding=1
> kern.polling.idle_poll=1
> kern.random.sys.harvest.ethernet=0
>
> and in /boot/loader.conf, add
>
> kern.hz="2000" 
>
> Also removing 
>
> options         ADAPTIVE_GIANT          # Giant mutex is adaptive.
>
> from the kernel helps a bit as well.
>
> with 
> kern.polling.idle_poll=1
>
> your load avg will be messed up but it should help performance a bit.
>
> As for firewall rules, things really seem to fall down performance
> wise, as compared to RELENG_4.  I havent found a way to improve that
> performance.... However, on the plus side, an extra core does seem to
> help a bit with the box remaining responsive.  For NICs, stay with em
> or bge nics for now in RELENG_6
>
> I have some misc test results at http://www.tancsa.com/blast.html
>
> 	---Mike
> --------------------------------------------------------
> Mike Tancsa, Sentex communications http://www.sentex.net
> Providing Internet Access since 1994
> mike at sentex.net, (http://www.tancsa.com)
>
>