Summary:ASTERISK-01486: ever increasing jitter buffer with iax2 -head
Reporter:zoa (zoa)Labels:
Date Opened:2004-04-28 11:19:28Date Closed:2011-06-07 14:10:27
Versions:Frequency of
Description:i just tried to put some load on the jitter buffer and it looks like its growing way to big.... a lot bigger than the jitter itself.

(Like 4 seconds big) and it seems to just keep increasing...

I'll try to give more info in the next days...
Comments:By: zoa (zoa) 2004-04-28 11:27:11

x.x.x.x        16394/16402  00009/00013  00088ms  167384ms  GSM -> looks like extreme amount of ms is still possible in cvs -head too.... no idea where it comes from

By: Brian West (bkw918) 2004-04-30 00:36:13

I don't think that is correct I think its a but in how it calcs ... something is slipping up somewhere... does the audio sound fine?

By: zoa (zoa) 2004-04-30 07:38:03

From doing a lot of iax2 show channels, i have the impression something has to be wrong with the received 'ringing' packets. (pre voice)

During ringing, i normally see 0ms delay with a normal jitter and as soon as the audio starts i see the normal delay, however sometimes during ringing the jitter display shows insane values.

Audio starts getting bad when the delay and reported jitter is getting too high.

I don't think you'll notice something when you are using a jitter buffer with few simultaneous calls and a small jitter buffer.

When a jitter buffer increases, does it increase for all calls ?
if so, when i have a trunk with 30 calls, and never have 0 simultaneous calls, there is no way the jitter buffer will ever be reset to the original value.

I also tried to use the dropcount option and the excess jitter buffer value in iax.conf, but delay still went skyhigh.

By: twisted (twisted) 2004-05-02 22:13:58

Unable to reproduce your results zoa.   the jitter obviously wasn't steady in my testing, but it didn't ever start increasing to insaine values, and even when the jitter went up higher, it was related to the delay in the call, and as the delay went down, the jitter acted accordingly.

By: zoa (zoa) 2004-05-03 04:35:31

did you try with a higher number of (trunked) calls twisted ?

As soon as i have some time i'll try to reproduce it again and find the exact circumstances.

By: Brian West (bkw918) 2004-05-04 21:43:45

When you can reproduce this please reopen. :)