[Asterisk-Users] Analyzer for Milliwatt

[Matt Roth]

Andrew Kohlsmith wrote:

>What is being discussed here is basically what I was planning on doing for an 
>automatic VOIP quality check.  Using miliwatt and analyzing it for 
>pop/jitter/etc as well as sending other known waveforms and comparing what 
>was received to what was expected and coming up with some "quality" number 
>which would be fed back to the dialplan to adjust the least-cost routing 
>paths.  Essentially come up with a "least cost but still good quality" 
>routing.  :-)
>
>I've done absolutely nothing other than a little research and a lot of 
>thinking about how to do it though.  I did some research on digital click/pop 
>removal for records as a way to detect poor quality, and then also some 
>monkeying around with coppice's excellent DSP routines in spandsp.
>
>-A.
>
Andrew,

This sounds like a programming project.  Something like a stripped down 
softphone (or possibly a plugin to an existing phone) with the ability 
to analyze the Milliwatt signal for variations/quality problems.  The 
ability to analyze other known waveforms would add a lot of value.

I suggest proposing your ideas to the -dev list or #asterisk-dev on 
FreeNode.  Someone else (I can't recall who) is working with SIPP in 
order to get it to pass the full RTP stream, instead of just the SIP 
signaling.  I believe that analyzing the quality of the RTP stream is 
still an open issue.  If it could be handled on a 1-to-1 basis by the 
call endpoints, it sounds like an elegant and scalable solution.

Currently, testing the scalability of an Asterisk system is a bit of a 
black art.  We did some work with an Abacus 5000 
<http://www.spirentcom.com/analysis/product_set.cfm?PS=73&PL=34&wt=2>, 
but they have a couple of significant drawbacks.  It was capable of 
originating and terminating hundreds of SIP calls, but it could only do 
audio quality analysis on up to 64 of them.  It is also a VERY expensive 
piece of equipment.

I'm very interested in your project, because our production system will 
push the vertical scalability of Asterisk.  So far we've handled 100 
concurrent calls with digital recording on a single server in a live 
environment with no quality issues, but the number of calls is going to 
increase to the 400-500 range as we add clients to the box.  The ability 
to test the results of the increased number of calls prior to going live 
could save me a LOT of headaches.  As such, your project is of 
significant value to myself as well as the community at large.  Please 
pursue it with the development community, and don't hesitate to contact 
me if needed.

Matthew Roth
InterMedia Marketing Solutions
Software Engineer and Systems Developer