Thanks again.
These devices could use 8 kbps audio, but they were using G.722 at 48 kbps in this case and advertise on their status screens that their audio bit rate is constant. Their sales people say they don't use silence detection. Other similar competive products also say their audio bit rate is fairly constant when they are connected to each other.
Their video data rates are reported as varying far more widely, as you mentioned, going down almost to 0 when no one moves on either end. In a "384" call these video devices report that they are sending anywhere from 320 kbps to 270 kbps when people are moving around on each end, and the data rate field updates about once every two seconds.
I have an Ethereal trace that I can compare to the ntop reports, and that should also help.
As for reading the C code, it does look like I can read it a bit. I will try to contribute what I learn by writing it up. I guess I would send it to you or post it somewhere. It may be interesting to others who run video over IP or voice over IP or similar stuff.
-----Original Message-----
From: Burton Strauss <[EMAIL PROTECTED]>
To: [email protected]
Sent: Fri, 15 Apr 2005 14:38:11 -0500
Subject: RE: [Ntop] Definition of "throughput" in Top Hosts Sent and TopHostsReceived
As to docs, nope - read the code, learn how things work, write documentation. That's a contribution to the project that ANYONE can make (it's really not that hard to read C code at a simplistic understanding level).
Yes, throughput measurement is done every minute - there's a clock routine in traffic.c, updateDeviceThpt(). The piece that walks through each host looking for the top three begins like this:
for(el=getFirstHost(deviceToUpdate); el != NULL; el = getNextHost(deviceToUpdate, el)) {
if(!isFcHost (el) && broadcastHost(el))
continue;
if(!isFcHost (el) && broadcastHost(el))
continue;
There are many reasons why ntop won't see traffic and thus won't report it - often this is due to switches, asymetric routing, etc. Or ntop could just be losing packets.
However, there's another possibility with audio and video - which is compression. While the rate required to support the channel may be 384Kbps, that doesn't mean the actual bandwidth isn't compressed way down. I know less about video, but here's something to think about:
Full 0-3000Hz audio (spoken word, i.e. telephone) takes 64Kbps. But there are codecs which run as low as 8Kbps. There are good sounding codecs in the 24-32Kbps range. Add into that some simple noise suppression (drop anything of a whisper or below) and then don't bother transmitting 'silence' and the observed bandwidth is a lot less.
Could something similar be happening with your video con? Don't know - that would depend on the equipment, compression, etc. But frame to frame, most video has very little changes - and so a difference based transmission protocol uses a lot less bandwidth.
-----Burton
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED]
Sent: Friday, April 15, 2005 2:14 PM
To: [email protected]
Subject: Re: [Ntop] Definition of "throughput" in Top Hosts Sent and TopHostsReceived
Thank you for the answer. Here is another question:
A and B in this case were video conferencing systems whose software applications say that they were sending an average of 384 kbps back and forth to each other for about 20 minutes.
Nevertheless the "A" sight which was the top sender during one minute of our observation was not the top sender in any of the minutes before or after that. In fact, the top sender during those other minutes was often much lower than 384 kbps. In fact, the top sender of the minute was often lower than 64 kbps and the video conferencing applications are fairly consistent abbout sending 64 kbps of audio, whatever the burst rate of the video may or may not be.
So, it seems that the video conferencing applications were either not sending data, or ntop was not looking for it for some reason. We had people in front of the video conferencing systems who did not report that the video or audio stopped at any point, so I think the video and audio data were being sent.
My question then is why doesn't ntop keep reporting the video conference IP addresses as top senders, or top receivers?
Does ntop report the top average over a minute?
Thanks, I hope this question(s) makes sense.
Is there are document that explains this ntop section in more detail?
<snip />
_______________________________________________ Ntop mailing list [email protected] http://listgateway.unipi.it/mailman/listinfo/ntop
_______________________________________________ Ntop mailing list [email protected] http://listgateway.unipi.it/mailman/listinfo/ntop
