[ previous ] [ next ] [ threads ]
 
 From:  David W. Hess <dwhess at banishedsouls dot org>
 To:  "Bob Young" <bob at lavamail dot net>
 Cc:  <m0n0wall at lists dot m0n0 dot ch>
 Subject:  Re: [m0n0wall] VoIP voice quality surprise
 Date:  Sun, 06 Aug 2006 16:32:31 -0500
On Sun, 6 Aug 2006 13:22:34 -0400, you wrote:

>I like the idea of using the ping command as you test Monowall.  I'll keep
>that in mind as I test.

I was more interested in low latency at the time so PING was very appropriate.
Worst case latency with no traffic shaping was more then 1500 milliseconds. With
traffic shaping it was from 400 to 800 milliseconds which would probably be fine
for VoIP since there should be no packet loss.  Using a separate set of queues
for low latency traffic got it down to 20 to 40 milliseconds.

Of course, being an engineering type I later wrote a script to compute standard
deviations and run other statistical tests to prove to myself that I was getting
valid data.

>On the other hand my PAP2 ATA must be setting TOS low delay flags, cause I
>have that checked off for my outgoing data.  So my outgoing rule matches the
>data my ATA is sending out, and so my VoIP outgoing pipe is being used.

I have not checked to see what my VoIP client sets the flags to.  I just use its
internal IP address to classify the packets.  I thought about placing it off of
one of the OPT interfaces but currently there is no advantage to doing so.

>Wow.  This stuff is really tedious...lol.  And interesting. 
>
>It is so neat that we all can get on here and help each other out.  (My wife
>says I need a lot of help...but that's a different matter...lol...Just
>kidding).

I do not post much here since my BSD and Monowall knowledge is fairly limited
compared to most who frequent this list.  In this case however I had specific
experience with the situation you were describing.  I am glad I could help.