On Oct 23, 2014, at 1:13 PM, Adam Thompson <[email protected]> wrote:

> On 14-10-23 03:06 PM, Chris L wrote:
>>> We don't release the tuning info, and, incredibly, a couple people a month 
>>> write in demanding it.
>> Does this mean there’s a special, hardware-specific version of pfSense (or a 
>> package or ?) or is the tuning in the hardware itself?
> 
> AFAIK it's the same software (plus or minus some logo and CSS changes? not 
> 100% sure...), but with different sysctl values precisely (in theory) matched 
> to the hardware it's running on.  I would imagine they also ensure all the 
> BIOS settings are set appropriately, IRQs are distributed appropriately, etc.
> 
> If you spent a few weeks testing the crap out of your own system, you'd be 
> able to figure out the precise values that maximized throughput for your 
> hardware, too.
> Note that the precise values that work for any particular piece of hardware 
> are unlikely to be precisely ideal for any other particular piece of 
> hardware... so even copying exactly what Netgate provides on *their* system 
> onto yours doesn't guarantee optimal performance.
> 
> Besides, given what Jim just said, do you really think he's going to answer 
> your question? ;-)
> The value-add is technically in the labour, but the "secret sauce" is knowing 
> precisely where to direct that labour to maximize the value to his paying 
> customers.
> The rest of us get enough value from the software as it is.
> 

I’m not asking what the changes are - I’m asking if these boxes require a 
special version of pfSense for maximum performance.

I am considering some C2758s and I’m curious.  I have another APU4 on its way 
to me as we speak.

If it’s just sysctl values then it’s not possible to keep it secret.  sysctl 
-a, sysctl -a, diff

If it’s a custom kernel, etc, then I have to take waiting for netgate to issue 
patches into consideration.  Now and in the future.


_______________________________________________
List mailing list
[email protected]
https://lists.pfsense.org/mailman/listinfo/list

Reply via email to