Carl Weinschenk spoke with Rich Watson, director of product management, DiVitas Networks. DiVitas recently partnered with AirMagnet to create enterprise voice over Wi-Fi platforms.
Weinschenk: What are the main challenges of running voice over a wireless LAN?
Watson: The first is QoS. If you look at the history, we've come a long way from the standards and industry perspectives from the early days starting with proprietary QoS to standards-defined QoS. For wireless VoWLAN, all is in place from a standards perspective. The missing piece from a commercial availability solution today is vendor compliance with the new standard, 802.11e, which is expressed by the Wi-Fi Alliance as WME [Wi-Fi Multimedia Extension]. The real question for QoS concerns compliance of the Wi-Fi infrastructure vendors ... and, more importantly, the support of handset vendors in complying with that standard. Once you've got QoS under your belt, one of the real struggles to do with VoIP is how the clients manage roaming. Specifically, to support good voice quality over WLANs, roams from access point to access points have to be very agile. You can apply the term "preemptive." For example, some Wi-Fi devices on the market today, even from major handset providers, have poor capabilities for roaming between access points. And what happens is that it gets translated to the user and the voice quality suffers or, in worse cases, the call is dropped. As long as the device is attached to [a single] access point, there is perfect QoS, the voice quality is really good. When you go to roam in today's world, that's where most products struggle. What has to happen from an industry standpoint is that vendors of the mobile handsets really have to architect Wi-Fi drivers to be voice-enabled.
Weinschenk: What happens if the system's voice-enablement is poor?
Watson: In the worst case scenario, a driver may hang on to an access point until it loses connectivity, at which point the driver goes through a discovery operation to discover a new access point. It has to associate and pass security interchanges with the access point before starting the conversation again. [In such cases] the latency or time window can be a half-second or longer. If it is longer, the call will probably drop. The bottom line is, again, the vendor, when writing the Wi-Fi driver, needs to architect with voice in mind. Having said that, they need to be looking ahead, they need to be aggressive and able to identify potential access point candidates. When one is walking through a building, as the signal degrades, the driver needs to be looking at other access point candidates it needs to be probing different channels, it needs to be aggressively identifying new access point candidates. When the [signal from the] existing access point goes below a certain point, it must make an aggressive or preemptive roam with the minimum of disruption to the call flow.
Weinschenk: Where is the industry in the continuum from totally proprietary systems to those that are fully interoperable and are able to fast roam?
Watson: From a commercial interoperable standpoint, I think we're really not there yet. I cannot go out and buy with high confidence a system today that will allow me to do fast roaming. Certain vendors do a very good job, others are in the learning curve. Specifically, if I'm a customer looking to deploy wireless VoIP, I want to understand the support [I'll get] from a voice perspective from the WLAN infrastructure provider and the kind of support [I'll get] from the handset provider. Typically, they are totally different vendors. It's mixing vendor A and B. For example, Meru Networks has specifically designed its system to be voice-optimized. If the customer has selected that vendor, they have much broader options in terms of selecting mobile handsets. [There are] handset vendors who have done a very good job of addressing this problem. I think both the handset industry and the infrastructure industry now are acutely aware of the possibilities of wireless VoIP.