[Asterisk-Users] maximum concurrent conference peers in asterisk

trixter aka Bret McDanel trixter at 0xdecafbad.com
Wed Nov 9 02:16:01 MST 2005


On Wed, 2005-11-09 at 21:32 +1300, Matt Riddell wrote:
> nr k wrote:
> > hi generally we describe the bandwidth in kilobits per
> > second only.
> 
> Cool, just checking, it seemed pretty low.
> 
> According to http://www.voip-info.org/wiki-Asterisk+bandwidth+iax2 you should
> be able to do 4 calls with g729.
> 

Bandwidth is a tricky issue.  You have your IP + UDP + RTP + whatever
headers (iax2 combines stuff so potentially that skews this a bit) but
something most often forgotten is link layer framing.

Take ATM (DSL uses ATM as do many other links).  ATM transmitted data is
chopped up into 53 byte cells.  Each cell has a 5 byte header.  This
leaves 48 bytes for payload per ATM cell.  Lets say your total packet IP
header on down is 80 bytes.  This means that on the ATM layer you have 2
ATM cells with 16 bytes of padding.  This is really only 83% efficient
network wise.  This is per RTP packet.

By adjusting your sample size you can try to fill the cell completly so
you dont waste extra bandwidth on padding (ATM cells can contain no more
than 1 packet and they are padded to fill the cell.  so every 48 bytes
of payload is another cell).  You dont want your sample size too small
however because that causes more IP overhead, too large and it can
degrade call quality (imagine a 30ms jitter buffer with 30ms sample
sizes, that means only 1 packet goes in the jitter buffer, with only one
packet you have the effect of no buffer at all, reordering packets is
impossible, delayed packets cant be normalized timewise, etc).  

Its a really fine balance and something you should consider if you
really want to tune your VoIP to your network.  Obviously once its
handed off to another network it becomes hard to create packets tuned
for a network you dont control, but its fair to assume that the majority
of backbone providers are doing ATM so by tweaking this you may find
that your voice traffic works better over the net at large too ...  YMMV

I didnt pay attention to what type of link the 64Kbps links were (I dont
think it was specified initially) so I dont know what framing is used,
but this is something to consider.  

By not paying attention to this fine detail you can waste a lot of
bandwidth then wonder why you start to have lossy performance when raw
bandwidth meters suggest you shouldnt have any loss.

This was something that I presented to the Sacramento Asterisk Users
Group last friday, although my power point presentation doesnt give the
subject the coverage it needs, most of that was audible.


-- 
Trixter http://www.0xdecafbad.com     Bret McDanel
UK +44 870 340 4605   Germany +49 801 777 555 3402
US +1 360 207 0479 or +1 516 687 5200
FreeWorldDialup: 635378
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.digium.com/pipermail/asterisk-users/attachments/20051109/62af04ff/attachment.pgp


More information about the asterisk-users mailing list