[asterisk-dev] IAX2 performance (was: Asterisk 1.4.20-rc1 Now Available)

[Tim Panton]

On 5 May 2008, at 21:21, Russell Bryant wrote:

> Tim Panton wrote:
>> Yep, it is a hack, but it might be worth considering instead of the
>> effort of trying to backport
>> the ast2obj fix, which is (of course) a much better fix.
>
> By the way, while we're on the subject of these changes, I figured I  
> would share
> the performance difference that I measured in brief testing.
>
> Testing between the code that included the security fix, and code  
> that included
> my performance enhancements yielded an increase in concurrent IAX2  
> channels of
> about 3600%.  :)
>
> However, in a more meaningful test, I went back to the code right  
> before the
> security fix.  Testing between that and the code with my performance
> enhancements yielded an increase of about 55% in concurrent IAX2  
> channels.
>
> I wonder what IAX2 vs. SIP performance looks like now ...
>

Nearer parity I guess.

Was that on a single or multiprocessor box ?
One of the problems with the old code was that it locked the list
whilst it it searched it, meaning that multi core was not going to be  
used to
best effect.

Tim.