[asterisk-dev] rtp scalability improvement...

[Steve Underwood]

Sergey Kuznetsov wrote:

> Roy Sigurd Karlsbakk wrote:
>
>>> PS: As of number of packets sent thru computer:
>>>
>>> 50 packets per second * 400 channels * 2 in/out = 40 000 packets per 
>>> second.
>>>
>>> Performance dies not because of system calls, but because 20 000
>>> interruptions/sec happened at that moment.
>>> It's named IRQ poisoning. Because IRQ switching is longer that standard
>>> task switching.
>>> <snip/>
>>
>>
>> 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
>>
> You are right about smart Intel network cards, but there is not that 
> much of them, and you have to know which one works well and which one 
> is not.
> Lots of network cards still generate one IRQ for one packet descriptor.

Only the very old or very nastiest (e.g. Realtek) do that. Interrupt 
mitigation has been mainstream for several years. Pretty much from the 
Tulip (and all its clones) forward, mitigation has been used, though 
some implementations are smarter than others.

Steve