[asterisk-users] Internal timing under load is critical ?

[Steve Edwards]

Please don't top post.

Please keep the thread only on the list.

> On Thursday, July 31, 2014 12:16 AM, Steve Edwards 
> <asterisk.org at sedwards.com> wrote:
> 
> I'm running Asterisk 11.11.0 on an HP ProLiant DL320e Gen8 E3-1240v2.
> 
> 1,300 calls with no audio issues.

On Wed, 30 Jul 2014, babak wrote:

> 1300 calls include playback voices ?

The test scenario was for the first server to originate calls (via call 
files) to the second server and then 'playback()' a long file. The second 
server would answer the call and then 'playback()' a long file. Audio was 
flowing in each direction.

Bandwidth was observed using 'iftop' as being in the 70mb to 80mb range in 
each direction (if I remember correctly).

I placed calls from a handset to confirm audio quality.

> which timing module you are using: res_timing_timerfd.so or 
> res_timing_kqueue.so or res_timing_dahdi.sores_timing_pthread.so

I used res_timing_timerfd.so. I'm finally making the leap from 1.2 to the 
current decade :) I read somewhere that this was the timer to use and it 
seems to be working fine for me.

I don't think the cores got much over 20% to 30% busy.

Various failures were observed on the console from running out of file 
descriptors. This was on a stock CentOS 6.5 install with no tweaks to bump 
up the max file descriptors.

The client only asked for 500 simultaneous calls so no further testing was 
done.

-- 
Thanks in advance,
-------------------------------------------------------------------------
Steve Edwards       sedwards at sedwards.com      Voice: +1-760-468-3867 PST
Newline                                              Fax: +1-760-731-3000