[asterisk-bugs] [Asterisk 0012708]: Dead air between answer and packet2packet bridge

[noreply at bugs.digium.com]

A NOTE has been added to this issue. 
====================================================================== 
http://bugs.digium.com/view.php?id=12708 
====================================================================== 
Reported By:                kactus
Assigned To:                
====================================================================== 
Project:                    Asterisk
Issue ID:                   12708
Category:                   Core/RTP
Reproducibility:            always
Severity:                   major
Priority:                   normal
Status:                     new
Asterisk Version:           1.6.0-beta8 
SVN Branch (only for SVN checkouts, not tarball releases): N/A 
SVN Revision (number only!):  
Disclaimer on File?:        N/A 
Request Review:              
====================================================================== 
Date Submitted:             05-22-2008 20:26 CDT
Last Modified:              06-24-2008 04:50 CDT
====================================================================== 
Summary:                    Dead air between answer and packet2packet bridge
Description: 
Hi we have been testing asterisk 1.6 extensively as we intend to replace
our long in the tooth 1.2 box that acts as our gateway between our offices
and the switched telephony network.

Asterisk 1.6 talks to directly to a cisco call gateway via sip which talks
to the out side world via PRI

One issue that we have noticed repeatedly is that there is a large delay
between when a call is answered and when voice traffic actually flows. The
delay is also asymmetrical and of the scope of about 2 seconds. This is
very noticeable as calling someone generally misses the entire greeting.

Call flow essentially goes like this:
start call -> ringing -> answered (other party start talking “welcome to
company this is Cameron”) -> their voice flows 2 seconds later and we
hear “ameron”

If we talk they can't here anything either at the beginning.

I have been mainly testing this with a snom 190 (have also tried sp962)
connected via sip to the 1.6 box (over nat).

We have also tested this by passing the voice out to one of the larger
voice providers (who also use cisco equipment) and they have stated time
and time again that it is not their end. Both Cisco gateways run
unauthenticated accepting calls from particular ips automatically.

RTP debug information is attached (RTP stats attached to bottom of it.)

Please let me know if you need anything else. We have run this on two 1.6
boxes one running beta 8 the other running beta 9.

====================================================================== 

---------------------------------------------------------------------- 
 kactus - 06-24-08 04:50  
---------------------------------------------------------------------- 
Hello been playing around with this a little more to maybe make it less
obscure an issue. 

I've created some test cases to maybe highlight the problem and so maybe
someone can suggest a further course of action/testing/debugging.

We have a couple of servers

A1.2 - old asterisk  1.2.9.1 current core of our network that we would
ultimately like to replace. It has a digium card within that talks straight
out one of our PRIs (PRI1). It has old hardware and has been running in
place since 2004 or so.
ATB - trixbox 2.6.0.0 office pbx
A1.6 - New asterisk 1.6beta9 machine which is configured to talk sip out
to our local cisco call gateway with a PRI wic (PRIcisco). This is by far
the beefiest box.

Psip - Snom 190 test phone that talks sip
Ppstn- standard pstn phone as benchmark

We are calling a local isp with a three cylable name and their greeting
kicks off as soon as their pbx picks up. Their greeting is "Welcome to ABC
support. Please note that for quality ...."

The greeting kicks in so quick that Ppstn sounds like it drops the W ie
“elcome to ABC ....” this is our bench mark against all others and I
called it from each system several time to ascertain there was no
variability.

Next we tested the setup as it is currently for the office.

Psip - SIP -> ATB – IAX2 -> A1.2 -> PRI1 = “elcome to ABC ...” ie
this has no perceivable delay relative to the pstn line.

Next we swapped ATB with A1.6

Psip - SIP -> A1.6 – IAX2 -> A1.2 -> PRI1 = “(t)o ABC support. Please
note....” ie it dropped the "welcome" and the “t” off “to”

Finally we removed the A1.2 box from the chain and the result was:

Psip - SIP -> A1.6 – SIP -> PRIcisco = “ort. Please note...” which
was around 2 and a bit seconds of dead air before voice flowed.

It seems the more that A1.6 has to do the longer the delay.

What tests would you suggest I try to get around this or what debug
information would be useful. I’m afraid I am unable to run a recent svn
due to the bug lodged as 0012508. 

I also thought it may be due to the dont_optimize compile flag so reran
the tests with dont_optimize unticked with no difference.

All the best - Kactus 

Issue History 
Date Modified   Username       Field                    Change               
====================================================================== 
06-24-08 04:50  kactus         Note Added: 0089139                          
======================================================================