Discussion:
[cisco-voip] Low IP Phone MOS Values
Andrew
2009-07-08 15:57:24 UTC
Permalink
Dear All,
I am currently troubleshooting issues with very low MOS values being
displayed on the IP Phone statistics. While the customer isn’t noticing
any issues with voice quality, I’ve been asked nevertheless to attempt
to resolve the issue.
There are around 20 IP Phones (7940/7961) in use on the site, and at any
given time 2 -6 of them can be displaying low values, usually the same
culprits every time. The phones are using the G711 codec and the
switches are connected (with 1 exception) via gigabit links.
The interface counters on the phones and the switches between the
phones, gateway and call manager (retracing the entire path) don’t have
any more than a normal volume of errors (>1%). QOS is configured on all
interfaces.
We made test calls both internally and externally with 5 separate phones
attached to each of the switches, and every time the MOS results were
between 4.48 and 4.5.
Values for the phones with issues look like:

MOS LQK 2.0000 Avg MOS LQK 2.2219 Min MOS LQK 2.0000 Max MOS LQK 4.1779
MOS LQK Version0.95
Cumulative Conceal Ratio 0.5963 Interval Conceal Ratio 0.0860 Max
Conceal Ratio 1.0000
Conceal Secs 140 Severely Conceal Secs 133
Latency 0 Max. Jitter 19

However at some point the values are going back to the same low values.
We enabled CMRs for the calls to determine if specific calls were
showing the poor values, however for all of the calls being logged the
media info is still carrying only the values “null”.

Does anyone have any advice on how we could try an resolve this issue,
or if not the MOS issue than the null values for the CMR records?

Thanks for any help anyone can offer,

Andrew Cook
Jason Burns
2009-07-08 22:32:26 UTC
Permalink
Andrew,
I'd put Wireshark on a PC behind one or more of the phones reporting the
problem. It can give you some pretty good RTP statistics regarding delay and
jitter. You can see how many RTP packets arrived in more or less than the 20
msec expected gap between packets, and how many might have been missed
altogether.

A good stream will show packets coming in at even 20 msec intervals with
consecutive sequence numbers.

Right click on the RTP packet in Wireshark and go to something like "Stream
Analysis" or highlight the packet and in the top menu go to "Statistics ->
RTP -> Stream Analysis".

Look for out of order sequence numbers, gaps bigger or smaller than 20
msecs, or big gaps in sequence numbers. If you can spot a particular
direction of the stream that looks goofed up you can identify all of the
ingress and egress interfaces between the two endpoints in the call and look
closer.

PhoneA -- Switch 1 -- Switch 2 -- Router 1 --- Router 2 -- Switch 3 - Switch
4 -- Phone B

Start with captures at Phone A and Phone B. If you spot problems, bisect the
link and capture between Router 1 and Router 2 to see which half is
introducing problems. Repeat.

If on the other hand you don't see any problem in the packet capture, then
maybe the phones are calculating something wrong.

For the Null CMR values, make sure you reset the phones. See if they start
populating CMR data after that.
Post by Andrew
Dear All,
I am currently troubleshooting issues with very low MOS values being
displayed on the IP Phone statistics. While the customer isn’t noticing any
issues with voice quality, I’ve been asked nevertheless to attempt to
resolve the issue.
There are around 20 IP Phones (7940/7961) in use on the site, and at any
given time 2 -6 of them can be displaying low values, usually the same
culprits every time. The phones are using the G711 codec and the switches
are connected (with 1 exception) via gigabit links.
The interface counters on the phones and the switches between the phones,
gateway and call manager (retracing the entire path) don’t have any more
than a normal volume of errors (>1%). QOS is configured on all interfaces.
We made test calls both internally and externally with 5 separate phones
attached to each of the switches, and every time the MOS results were
between 4.48 and 4.5.
MOS LQK 2.0000 Avg MOS LQK 2.2219 Min MOS LQK 2.0000 Max MOS LQK 4.1779
MOS LQK Version0.95
Cumulative Conceal Ratio 0.5963 Interval Conceal Ratio 0.0860 Max Conceal
Ratio 1.0000
Conceal Secs 140 Severely Conceal Secs 133
Latency 0 Max. Jitter 19
However at some point the values are going back to the same low values. We
enabled CMRs for the calls to determine if specific calls were showing the
poor values, however for all of the calls being logged the media info is
still carrying only the values “null”.
Does anyone have any advice on how we could try an resolve this issue, or
if not the MOS issue than the null values for the CMR records?
Thanks for any help anyone can offer,
Andrew Cook
_______________________________________________
cisco-voip mailing list
https://puck.nether.net/mailman/listinfo/cisco-voip
Tim Smith
2009-07-08 23:29:55 UTC
Permalink
Hi Guys,

Not sure on your MOS scores. You could look at doing some testing with Cisco
IP SLA.

The Null values in CMR should be easy.

First of all - check your service parameters

1. Under CallManager service parameter - check "CallDetailRecordDiagnostic"
is set to always enabled regardless of CDR. (It is cluster wide so you only
need to do it on one server)

2. Go to CDR Analysis and Reporting Tool. Check the CDR loading schedule.
There is a tick box here that says "Load CDR only" it is on by default -
hence no CMR records are loaded.
Untick it. Save and then restart your CAR Web Scheduler service.

Then make some test calls and go back and check CDR / CMR.

Cheers,

Tim
Post by Andrew
Dear All,
I am currently troubleshooting issues with very low MOS values being
displayed on the IP Phone statistics. While the customer isn’t noticing any
issues with voice quality, I’ve been asked nevertheless to attempt to
resolve the issue.
There are around 20 IP Phones (7940/7961) in use on the site, and at any
given time 2 -6 of them can be displaying low values, usually the same
culprits every time. The phones are using the G711 codec and the switches
are connected (with 1 exception) via gigabit links.
The interface counters on the phones and the switches between the phones,
gateway and call manager (retracing the entire path) don’t have any more
than a normal volume of errors (>1%). QOS is configured on all interfaces.
We made test calls both internally and externally with 5 separate phones
attached to each of the switches, and every time the MOS results were
between 4.48 and 4.5.
MOS LQK 2.0000 Avg MOS LQK 2.2219 Min MOS LQK 2.0000 Max MOS LQK 4.1779
MOS LQK Version0.95
Cumulative Conceal Ratio 0.5963 Interval Conceal Ratio 0.0860 Max Conceal
Ratio 1.0000
Conceal Secs 140 Severely Conceal Secs 133
Latency 0 Max. Jitter 19
However at some point the values are going back to the same low values. We
enabled CMRs for the calls to determine if specific calls were showing the
poor values, however for all of the calls being logged the media info is
still carrying only the values “null”.
Does anyone have any advice on how we could try an resolve this issue, or
if not the MOS issue than the null values for the CMR records?
Thanks for any help anyone can offer,
Andrew Cook
_______________________________________________
cisco-voip mailing list
https://puck.nether.net/mailman/listinfo/cisco-voip
Loading...