How the Big Guys Support VoIP Through Mesh Network Transitions

Seamless transition between mesh network sites is a tough challenge for data access. At Spring 2006 VON, the even greater challenge of VoIP transitions was addressed:

Stu Milberg from Vivox moderated a panel that included:

• Tim Gore, Lucent
• Matt Holdredge, Strix Systems
• A well-spoken integrator that deploys BelAir networks (unfortunately, I didn't get his name, and he is not in the program)

Gore had the most useful slide deck among the presenters. He started off with the perspective that Lucent doesn't develop any WiFi products, they act as a systems integrator. And his neat turn of phrase was that WiFi mesh turns hotspots into hot zones.

Our BelAir rep reminded everyone that there are many types of mesh networks: single radio shared (2-8 Mbps of throughput), dual radio shared (20 Mbps), and multi radio switched (75 Mbps). Of these, switched mesh allows the radio to have full use of spectrum, so it better supports voice. This is reminiscent of when cellular technologies were being launched, and how switching helped bandwidth/voice quality there too.

A total network approach to VoIP, as Gore defines it, balances an eye on the user experience with an understanding of network areas. In a network area, you have to consider both the access network and the core network, and you have to provide a data center. Seamless transitions should include both WiFi-to-WiFi and WiFi-to-cellular handoffs. And in the future, they'll need to provide interoperability between WiFi and non-802 networks. (gasp!)

There were some interesting statistics in Gore's talk - in particular, it was notable that 75% of mesh networks can't maintain optimal (although, what is optimal?) VoIP quality. These networks weren't engineered, and aren't managed, to support VoIP. Capacity is too low and network delays are too long.

If you do want to implement VoWiFi, then Gore suggests that you assess the network to ensure good QoS on both the LAN and the WiFi network. This is absolutely critical if you want acceptable voice quality, yet most people fail to take both measures. Gore shared some network usage analysis for the Olney muni wireless system that showed its dramatic adoption rate and increases in traffic per user.

Mesh architecture design, to be effective, must include call capacity, packet processing, latency, QoS, and mobility. You can even stream video over mesh networks (and some people are), assuming that the network is designed with IPTV in mind as an application. The characteristics of expected applications - bandwidth requirements, latency requirements, etc. - must clearly be identified in network design, and seeing what it takes to deliver effective IPTV support makes this very clear.

Note: The Bel Air and Strix speakers also provided plenty of information on their own products and showcase installs, but a recitation of features didn't seem too useful for the topic. Check their web sites if you want more details.

Tags: christine herron christine.net space jockeys mobile wifi  voip mesh network von

Comments

[Shannon Newton]

Nice Blog. Interesting thoughts on what it will take to make mesh networking work reliably with Video. One thing. When you say effective architecture must include latency, (I assume you mean design to reduce latency) does that mean some type of partnership between the various owners of the pieces of the mesh? If so, what form do you think this will come in? Some established standards (and by whom) or a simple consolidation of the industry?