[asterisk-dev] rtp scalability improvement...

[Sergey Kuznetsov]

Tim Panton wrote:
>> These IRQ storms are only happening on crappy network hardware. My 
>> testing was one with intel gigabit NICs with large buffers, 
>> effectively producing < 100 interrupts per second. kernel profiling 
>> showed time was indeed spent in system calls
>>
>
> Yet another reason IAX trunking wins, 50 channels in a trunk means 49 
> fewer packets,
> hence fewer context switches.
>
> Is there a halfway house here, with a kernel driver that just 
> aggregates a number of rtp
> packets and hands them all back in one delimited buffer?
>

In this case you will end-up only with Linux. Otherwise you have to have 
lots of driver versions for different OS kernels, and so on.


I prefer to do this stuff in user-land even if I hardcore system 
programmer, who likes to put everything into kernel =)


PS: I still wondering how Linux kernel network layer implemented. In 
some places it's brilliant, in some places just full of crap-load.
I like the VJ's idea of network channels a lot! Such a clean and not 
bloated realization.



All the Best!
Sergey.