Wireless Access

I've been digging through some Lync call reports of customers that reported poor call quality in the last month.  Most of the time the issue reported was letters and/or words are being dropped and sometimes the voice would sound robotic.  Often, the call report indicates that jitter was the number one issue causing MOS degradation, as well as some packet loss.  I'd like to have a better understanding of the correlation between jitter/packet loss as reported by voip client and what's happening on the wireless side.  Configuration wise, the wireless network is configured to follow all current Lync best practices, so I don't want to focus on that.  Also,we've poured through all the wireless metrics and for the most part things look great.  I want to focus more generally on how wireless issues cause VOIP problems.

So here are my questions:

1. What wireless metrics do you focus on when poor call quality is reported?  Channel utilization, retransmits, RTS failures, client health, etc?
2. Based on those metrics, what do you consider the thresholds to be?  SNR 30+, 5% retransmits, etc.
3. What wireless problems result in jitter as reported by the VOIP client?
4. What wireless problems result in packet loss as reported by the VOIP client?
5. How do client retransmits affect the communication of a voice call?  Will the client attempt to resend the frames or not since they're UDP?

In most of the research I've done on the wireless side, I'm not seeing any smoking guns.  Client health, AP capacity, etc all look great.  Sometimes I'll notice a small fluctuation in channel utilization and retransmits, but these are usually fairly low - around 5-10% for both.  For now, I feel a bit like I'm chasing ghosts here since nothing stands out as a problem on the wireless side.

Looking forward to hearing from others.

I guess the first question is what do you use for the LYNC clients?  Laptops or mobile devices?

Laptops.

Are the laptops stationary or mobile when the problem is occuring?  Are your LYNC servers in-house or is this an Office 365 deployment?

Laptops are stationary.

Servers are in house.

Hi thecompnerd,

I would be really interested to hear from those who have successfully measured sites with bad wireless voice calls with any of the wireless parameters as well (be it the channel utilization, retransmit etc). Kindly share your experience if any. 

Here is a few questions that is not clear for your network environment:

1) Does your network have proper (bidrectional) QoS enabled in the Wide Area Network (WAN) working?

- This will make sure that the problem that you are seeing is coming from the WAN or WLAN.

2) Are you able to split the number of call from the LYNC reporting (one for wired call and another for Wireless calls)?

- Being able to split the bad calls for a site for clients connected to Wired or Wireless will tell you if it's a problem isolated to Wired or Wireless or both.

3) Are your laptop clients configured to do the QoS marking using GPO policies so that the voice packets are being marked correctly when it leave the end device laptop?

- I won't even touch QoS until I have a good grip on how many good call and bad call for a site to see the effect of activating QoS. Maybe that's just me. 

HTan2,

1) Yes.  End-to-End QoS is configured/working.

2) Yes.  Call reports are done by site so we can narrow down if the problems are between sites or within a site.

3) Yes.  QoS policy is pushed down to all Windows devices.  Verified over the air tagging is working bi-directionally and tags DSCP mappings are correct when they hit the wired network.

I wanted to spend more time on the RF side of things since we know Lync and the network are configured as they should be.

Hi thecompnerd,

On Point 2,

- Did you split the reports further from within the site, how many good call are coming from the Wired on that site and how many are coming from the Wireless? And, are you seeing majority of the bad calls coming from the Wireless Connected Clients? 

- If you did see those coming from the Wireless Connected Clients, another thing to do is to dissect those bad calls according to the BSSID that is available on the LYNC reporting and see if it's isolated to one AP at the site or spread out across multiple different access points.

- Once you have identified the access points that is problematic, then you will need to drill into the calls to see what is source of the bad call for those calls (if it's due to jitter, packet loss or in some cases, it could be client high CPU, client using internal MIC etc). 

Hope that helps.