Comments on: Designing for Voice is a Big Deal!

By: Anonymous

Anonymous — Wed, 14 Dec 2011 23:33:00 +0000

Thanks for your comments. I agree that analysis is a requirement to really understand what is going on.

By: Jon Linton

Jon Linton — Wed, 14 Dec 2011 19:58:23 +0000

Keith,
First great write up and I share these same thoughts. I would also mention that it is necessary to verify how much VoWiFi traffic already exists. You may see all kinds of traffic categorized as VO queue like Facetime because of a firewall or core seeing RTSP traffic and marking it accordingly. I network analysis is absolutely necessary.

By: KeithRParsons

KeithRParsons — Sat, 29 May 2010 19:43:18 +0000

Doh! you're right of course

By: WiFiNut

WiFiNut — Sat, 29 May 2010 09:47:00 +0000

I think you mean milliseconds…right?

By: Leonardo Mezzanotti

Leonardo Mezzanotti — Sat, 29 May 2010 03:18:48 +0000

But WLAN retries occurs at micro seconds order, how can it be too late. I think it all depends on the WLAN traffic load.

By: KeithRParsons

KeithRParsons — Sat, 29 May 2010 01:06:16 +0000

Yes, you are correct that WLANs will retry on a failed packet. But it is too late… because of the real-time nature of VoIP, and the reason VoIP uses UDP, is because the time delay in retransmitting, will generate a 'useless' packet. Any voice packet that arrives out of sequence, or delayed, will be discarded from the Codec and not processed.

By: KeithRParsons

KeithRParsons — Sat, 29 May 2010 01:04:40 +0000

Thanks for the power asymmetry issue. You are correct. One more to add to my list…

By: WiFiNut

WiFiNut — Fri, 28 May 2010 04:17:13 +0000

Excellent piece Keith! I wanted to add also the power asymmetry issues that can sometimes rear its ugly head. My company deals with it all the time because of our device's relatively low transmit power. Such issues cause one way audio.

By: Leonardo Mezzanotti

Leonardo Mezzanotti — Fri, 28 May 2010 01:52:00 +0000

Congrats, very nice list, it's on my reference list.
You stated that "Achieving a 1% packet loss is highly unlikely given the physics of Wi-Fi". Don't you think that WLAN retries can make the UDP voice packets loss less than 1% or even 0% loss?
Check http://www.cwnp.com/bbpress/topic.php?id=5305#p...

By: WLW 010 – Designing for Voice / Tech Tools : Wireless LAN Professionals

WLW 010 – Designing for Voice / Tech Tools : Wireless LAN Professionals — Wed, 24 Mar 2010 07:04:25 +0000

[...] addition to the ideas and topics covered in a previous blog post on “Designing for Voice is a Big Deal!” – Chris and Keith discuss some additional ideas on how to design for Voice over [...]

By: SteveJWilliams

SteveJWilliams — Thu, 04 Mar 2010 03:53:45 +0000

Thanks for your comment about my blog and for the update to the Cisco document. About separating voice & data on different bands, or designing dense WLANs, budget is always a big factor.

By: KeithRParsons

KeithRParsons — Wed, 03 Mar 2010 20:09:20 +0000

Steve,

Thanks for the comments and the links. Some of those CIsco docs were a bit dated. Try their newer 7921 deployment guides for the latest http://www.cisco.com/en/US/docs/voice_ip_comm/c...

It is a very good idea to separate the Voice and Data onto different frequencies! (if you can)

By the way, I like what you are doing over on your site! http://blog.sudonetworks.com/

By: KeithRParsons

KeithRParsons — Wed, 03 Mar 2010 20:09:02 +0000

Marcus,

It is a little intimidating to have so many variables to deal with. Though it is possible... just frustrating when people think adding Voice to a Wireless Network is going to be 'easy'. Far from it.

One of the main issues -- some folks think you can just throw more APs at the situation... but that only exacerbates the co-channel interference and adds to the contention domains (collision domains)

If you can think of anything else to add to the list - send it over and I'll post the new items.

Thanks for your feedback

By: Marcus Burton

Marcus Burton — Wed, 03 Mar 2010 20:04:20 +0000

Very nice, Keith. Thanks for sharing your expertise here. It is amazing to think about all of these variables in one shot, and I'm sure this list could grow too.

By: SteveJWilliams

SteveJWilliams — Wed, 03 Mar 2010 07:44:48 +0000

Excellent post Keith, actually it could be used as a checklist for designing VoWLANs. Here is some Cisco pointers I found: with Cisco 7921G deployments there can be up to 27 phones on call per ap/channel at 24 Mbps or higher. Sources: http://tinyurl.com/yav2og9 and http://tinyurl.com/5nk6a7 (page 31). An older Cisco document shows how to calculate the max number of phones for a 802.11b WLAN (8 calls per AP) source: http://tinyurl.com/yclw5pt. Don't know about other vendors. These are vendor figures only, not how-to's or best practices. What I haven't seen a lot so far (in my debutant WiFi career) is mixed deployments of voice and data, sharing the same APs. I've mainly seen data only and voice only WLANs. Is it a good practice to dedicate a particular band for voice (2.4GHz) and the other for data (5GHz) ? Are there other good practices for VoWLANs ? Like smaller cells or greater AP density (does that mean a complete separate set of APs for voice ?) What if we throw 802.11n in the equation ? For now, I rely on my AirMagnet Planner software and its what-if scenarios.