Wireshark-dev: [Wireshark-dev] Idea for Very powerful VOIP over WLAN Quality rtcp measurements
From: "Peter Klein" <[email protected]>
Date: Wed, 12 Nov 2008 16:50:02 +0100
Hi
 
I have an idea that I can make the proof of concept for but do not have the skills to follow through to something usefuls for others.
 
I'd like to test this with you if it would at all be possible to acheive in Wireshark, if anyone else thinks it would be valuable
and if someone has the skills to do it.
 
 
My setup is that I have a WLAN built on Cisco LWAPP Access Points, although this could be anything...
Over these I run Cisco 7921 WLAN phones, again, could be any Wireless WLAN phones.
 
The customer claim to have problems with these phones but it's really difficult to find out where the problem is.
 
I have used Wireshark to capture all traffic and filtered out all RTCP information. After this I have extracted the IP source and destinations for both the phones as well as the APs as well as the RTCP fraction number.
 
From this I have created this graph which takes some time to grasp but is very very powerful to be able to understand what is going on in the network.
 
What you can see is this:
Each bubble represents one RTCP report.
The color represents each phone (the IP-adress of the phone)
The X-axis represents the time of the packet (report)
The Y-axis represents on which AP the call is located. (IP-adress or wlan.da or something)
The size of the bubble represents the rtcp.ssrc.fraction (+1) so that we can see the quality of the call.
 
SO what I can read from this graph is that the phones that are represented by the Blue, White and Red bubbles problably have some issues in their calls. The phone with the Blue (16) stays on the same AP but gets more issues at 9:07:30 something, perhaps because the call from phone 6 (also blue unfortunately roams to the same AP and then back and forth.
 
We can also see that the redish call with phone 14 stays on the same base and is not affected that the yellow call roams to it in the middle of the graph.
 
Of course this does not say anything by itself but it really gives me a good place to start looking for where the problem lays and since I have the capture file I can easily dig down deeper into why certain phones do not behave.
 
 
 
 
So, what do I want then?
 
I would like for you, yes you! to respond if you feel that this would be useful to you, perhaps have some commenst but most of all if you find a good way of implementing this into Wireshark directly or if there is an easy way to export data to do this out of wireshark.
 
Today I have exported a PDML file, created a custom XSL that creates a much smaller XML file that I import into Excel. From there I still have a lot of manual labour to be able to produce the graph which makes it very cumbersome to use. But I think it should be very useful if it was easy to setup and I know that it can be done.
 
SO, if you have some comments, please do not hesitate to respond or comment to the list and perhaps we can create something all together.
 
/Peter Klein
TDC Sweden