[Freeswitch-users] CONFIG_HZ=1000 vs mod_timerfd -- preferred approach

Peter Olsson peter at olssononline.se
Thu Aug 29 14:54:29 MSD 2013 

I usually end up using Wireshark to trace the actual RTP stream. There are
many reasons for bad audio, timing is one, network's another one. I have no
experience of gsmopen endpoints though, so I really don't know how to take
that part any further.

/Peter


2013/8/29 Jayanth Acharya <jayachar88 at gmail.com>

> Thanks again Peter.
> So my theory attributing timer issue to the terrible audio quality goes
> away. Need to find what else could be the cause !!
>
>
> On Thu, Aug 29, 2013 at 2:36 PM, Peter Olsson <peter at olssononline.se>wrote:
>
>> Yes, it might be a typo. IIRC it should be usec for time_test. The small
>> sample of timer_test you provided seems ok to me, +/- 40 usec is probably
>> good enough.
>>
>> I think you're good to go :)
>>
>> /Peter
>>
>> 2013/8/29 Jayanth Acharya <jayachar88 at gmail.com>
>>
>>> On Thu, Aug 29, 2013 at 1:28 PM, Peter Olsson <peter at olssononline.se>wrote:
>>>
>>>>
>>>> I believe that timerfd is used automatically in the core if it detects
>>>> a system supporting it. However, I don't know how to really be sure.. :)
>>>> One thing though, if you just startup FS, if it doesn't try to do clock
>>>> calibration during startup, I think it is using timerfd stuff, so that
>>>> might be a way to "know"...
>>>>
>>>> Also, when running FS, you can try timer_test command (inside FS
>>>> console) and check out the results in there.
>>>>
>>>
>>> Thanks Peter.
>>>
>>> freeswitch at dabbian1> timer_test 10 3
>>> 2013-08-29 13:56:24.703028 [CONSOLE] mod_commands.c:827 Timer Test: 1
>>> sleep 10 10019
>>> 2013-08-29 13:56:24.712967 [CONSOLE] mod_commands.c:827 Timer Test: 2
>>> sleep 10 9930
>>> 2013-08-29 13:56:24.722996 [CONSOLE] mod_commands.c:827 Timer Test: 3
>>> sleep 10 10040
>>>
>>> Avg: 9.996ms Total Time: 29.989ms
>>>
>>> However,  for the "time_test" command in Mod_commands Wiki page -- I see
>>> that it assumes units as "microseconds" (not milliseconds, which is usually
>>> what we mean by "ms").. or is it a typo ??
>>>
>>> *... bad timer jitter is. It runs the test count times (default 10) and
>>> tries to sleep for mss microseconds. It returns the actual timer
>>> duration along with an average.
>>> *
>>>
>>> Anyhow, not sure how to interpret if the output of timer_test indicates
>>> a healthy resolution or not.
>>>
>>> /Peter
>>>>
>>>>
>>>> 2013/8/29 Jayanth Acharya <jayachar88 at gmail.com>
>>>>
>>>>> Thanks Jeff. Reading the mod_timerfd wiki page, I see that there are 2
>>>>> approaches to using timerfd,
>>>>>
>>>>> 1. Using the loadable module i.e. mod_timerfd (requiring me to build
>>>>> this module and then ensure it is loaded)
>>>>> 2. Using the support of timerfd now available in FS core itself
>>>>> (requiring me to just enable it in the switch.conf file)
>>>>>
>>>>> By default neither of those 2 seem to be true. I am on what as the Git
>>>>> HEAD last week - 22nd Aug 2013, so pretty recent source. However, it when I
>>>>> adopted the 2nd approach, i.e. enable timerfd use in FS core, by enabling
>>>>> the config in switch.conf, it didn't have the desired effect. Perhaps
>>>>> indicating that my problems might not be limited to just the timer
>>>>> resolution !!
>>>>>
>>>>> Using the method described here:
>>>>> http://www.advenage.com/topics/linux-timer-interrupt-frequency.php
>>>>>
>>>>> I tested and found that on my FS machine, the timer resolution
>>>>> available is already pretty good -- in the ~3KHz range, so shouldn't be an
>>>>> issue, if FS core is indeed correctly using timerfd.
>>>>>
>>>>> ---
>>>>> kernel timer interrupt frequency is approx. 3401 Hz
>>>>> jayachar at dabbian1:~$ ./timer_resolution_test
>>>>> kernel timer interrupt frequency is approx. 3401 Hz
>>>>> jayachar at dabbian1:~$ ./timer_resolution_test
>>>>> kernel timer interrupt frequency is approx. 3448 Hz
>>>>> jayachar at dabbian1:~$ ./timer_resolution_test
>>>>> kernel timer interrupt frequency is approx. 3021 Hz
>>>>> jayachar at dabbian1:~$ ./timer_resolution_test
>>>>> kernel timer interrupt frequency is approx. 3067 Hz
>>>>> ---
>>>>>
>>>>> I have already check that my kernel has the TIMERFD option enabled in
>>>>> the build, which explains (perhaps) the ~3KHz resolution I see.
>>>>>
>>>>> So the question now is, is FS really using the timerfd logic in it's
>>>>> core properly, or do I need to try the mod_timerfd method as well !!
>>>>> (That's a rhetorical question... I will go ahead and try anyway, disabling
>>>>> the FS core option to use timerfd, as per the Wiki).
>>>>>
>>>>>
>>>>>
>>>>> On Thu, Aug 29, 2013 at 11:21 AM, Jeff Leung <jleung at v10networks.ca>wrote:
>>>>>
>>>>>> If I can recall correctly, timerfd will be used automatically
>>>>>> regardless in recent versions of FreeSWITCH if it’s available.****
>>>>>>
>>>>>> ** **
>>>>>>
>>>>>> *From:* freeswitch-users-bounces at lists.freeswitch.org [mailto:
>>>>>> freeswitch-users-bounces at lists.freeswitch.org] *On Behalf Of *Jayanth
>>>>>> Acharya
>>>>>> *Sent:* Wednesday, August 28, 2013 9:35 PM
>>>>>> *To:* FreeSWITCH Users Help
>>>>>> *Subject:* [Freeswitch-users] CONFIG_HZ=1000 vs mod_timerfd --
>>>>>> preferred approach****
>>>>>>
>>>>>> ** **
>>>>>>
>>>>>> Assuming that one has full control over the kernel configuration and
>>>>>> FS build / config options, is there a preferred mechanism between the 2 for
>>>>>> letting FS have access to high resolution timers, i.e.****
>>>>>>
>>>>>>    - CONFIG_HZ=1000****
>>>>>>    - mod_timerfd****
>>>>>>
>>>>>> This was one step I missed in my FS setup, and I think it might
>>>>>> explain the extremely choppy, broken, garbled announcement I hear when I
>>>>>> call into my gsmopen EP via phone.****
>>>>>>
>>>>>> J****
>>>>>>
>>>>>>
>>>>>> _________________________________________________________________________
>>>>>> Professional FreeSWITCH Consulting Services:
>>>>>> consulting at freeswitch.org
>>>>>> http://www.freeswitchsolutions.com
>>>>>>
>>>>>> 
>>>>>> 
>>>>>>
>>>>>> Official FreeSWITCH Sites
>>>>>> http://www.freeswitch.org
>>>>>> http://wiki.freeswitch.org
>>>>>> http://www.cluecon.com
>>>>>>
>>>>>> FreeSWITCH-users mailing list
>>>>>> FreeSWITCH-users at lists.freeswitch.org
>>>>>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
>>>>>> UNSUBSCRIBE:
>>>>>> http://lists.freeswitch.org/mailman/options/freeswitch-users
>>>>>> http://www.freeswitch.org
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> _________________________________________________________________________
>>>>> Professional FreeSWITCH Consulting Services:
>>>>> consulting at freeswitch.org
>>>>> http://www.freeswitchsolutions.com
>>>>>
>>>>> 
>>>>> 
>>>>>
>>>>> Official FreeSWITCH Sites
>>>>> http://www.freeswitch.org
>>>>> http://wiki.freeswitch.org
>>>>> http://www.cluecon.com
>>>>>
>>>>> FreeSWITCH-users mailing list
>>>>> FreeSWITCH-users at lists.freeswitch.org
>>>>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
>>>>> UNSUBSCRIBE:
>>>>> http://lists.freeswitch.org/mailman/options/freeswitch-users
>>>>> http://www.freeswitch.org
>>>>>
>>>>>
>>>>
>>>>
>>>> _________________________________________________________________________
>>>> Professional FreeSWITCH Consulting Services:
>>>> consulting at freeswitch.org
>>>> http://www.freeswitchsolutions.com
>>>>
>>>> 
>>>> 
>>>>
>>>> Official FreeSWITCH Sites
>>>> http://www.freeswitch.org
>>>> http://wiki.freeswitch.org
>>>> http://www.cluecon.com
>>>>
>>>> FreeSWITCH-users mailing list
>>>> FreeSWITCH-users at lists.freeswitch.org
>>>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
>>>> UNSUBSCRIBE:
>>>> http://lists.freeswitch.org/mailman/options/freeswitch-users
>>>> http://www.freeswitch.org
>>>>
>>>>
>>>
>>> _________________________________________________________________________
>>> Professional FreeSWITCH Consulting Services:
>>> consulting at freeswitch.org
>>> http://www.freeswitchsolutions.com
>>>
>>> 
>>> 
>>>
>>> Official FreeSWITCH Sites
>>> http://www.freeswitch.org
>>> http://wiki.freeswitch.org
>>> http://www.cluecon.com
>>>
>>> FreeSWITCH-users mailing list
>>> FreeSWITCH-users at lists.freeswitch.org
>>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
>>> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
>>> http://www.freeswitch.org
>>>
>>>
>>
>> _________________________________________________________________________
>> Professional FreeSWITCH Consulting Services:
>> consulting at freeswitch.org
>> http://www.freeswitchsolutions.com
>>
>> 
>> 
>>
>> Official FreeSWITCH Sites
>> http://www.freeswitch.org
>> http://wiki.freeswitch.org
>> http://www.cluecon.com
>>
>> FreeSWITCH-users mailing list
>> FreeSWITCH-users at lists.freeswitch.org
>> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
>> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
>> http://www.freeswitch.org
>>
>>
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> 
> 
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://wiki.freeswitch.org
> http://www.cluecon.com
>
> FreeSWITCH-users mailing list
> FreeSWITCH-users at lists.freeswitch.org
> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
> http://www.freeswitch.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-users/attachments/20130829/0ad252c0/attachment-0001.html

Join us at ClueCon 2013 Aug 6-8, 2013
More information about the FreeSWITCH-users mailing list