Making VoFi Work
I spent the past couple of days at the annual Baltimore edition of the Wi-Fi Planet show, now officially known as Wi-Fi/VoWi-Fi Planet. For the record, I refuse to use VoWi-Fi as the Official Acronym for Voice over IP over Wi-Fi. VoFi is the absolute minimum number of syllables required to unambiguously get the point across. Who's with me on this? Write your congressperson today!
But seriously, the folks who put on the above-named-but-much-too-long-to-type-again conference are to be commended for recognizing the potential of VoFi, which is destined to become a core driver of the WLAN/Wi-Fi industry. I haven't written as much on this topic as I should have, and I will have more later this year on my reasoning as to why this prognostication is and shall be. But, for now, I want to report on a couple of interesting points made at the sessions I chaired on "Capacity and Quality Issues for VoWiFi and Mesh Networks as VoWiFi Platform." I also did my usual two-hour, drink-from-a-firehose tutorial, which is now designed for those with at least some experience in WLANs as opposed to rank beginners (I'm assuming there are none of the latter left, at least at conferences like this one). If you'd like a copy of these slides, just drop me a note at craig@farpointgroup.com.
The first conference session above really got to the heart of the matter with respect to adding voice to WLANs. Any single voice stream doesn't require much in the way of bandwidth - I like to use 200 Kbps to represent each full-duplex call, because it's a nice round number, but the actual requirement can be much less when voice coding and compression are applied. But voice does present a serious requirement for headroom, meaning that sufficient bandwidth must be reserved in the Wi-Fi network to assure that voice-bearing packets can be forwarded with minimal latency. End-to-end latency really can't be more than 30 milliseconds or so, or the voice quality will become unacceptable. So, add lots of calls to a Wi-Fi network, and capacity suddenly becomes a big concern.
The panel came up with a number of interesting points and suggestions. First of all, while there was some agreement that moving to 802.11a with its vastly greater spectrum availability would help, it was also agreed that end-to-end protocols are critical to success of any VoIP service, wireless or not. The special nature of wireless bandwidth, with its highly variable quality (often on a moment-to-moment basis), represents a significantly different challenge than one would see on wire. Managing APs centrally so as to avoid interference is important, but no one had a really good solution for the problem of external interference from other devices operating nearby, probably because there isn't one. RF Spectrum Management (RFSM), which I've written about before, is the best approach, but no mitigation technique is ever going to be perfect. There was also some discussion of security, but I think this is a red herring - after all, we're not terribly concerned about security on landlines or cell phones, so what's the big deal with VoFi security? Regardless, we have all of the 802.11 and Wi-Fi security mechanisms available to us, and we can even apply higher-level end-to-end security protocols if the application so merits.
The mesh session was even more interesting. One of the key challenges with meshes is latency as packets are relayed from one node to another through the mesh, and, as I mentioned above, latency is something that must be controlled in any time-bounded situation like voice. As I noted previously, mesh protocols can go a long way towards addressing this problem. But ultimately, capacity is also the core challenge here, just as it is with more traditional wireless LANs. You can add more capacity by adding additional infrastructure mesh nodes (APs), and by adding more backhaul (connections between infrastructure nodes and external networks via wire, dedicated Wi-Fi links, or perhaps another wireless technology like WiMAX). But most of the panelists voted for yet another solution - multi-radio mesh nodes. As you know, adding more nodes (or more radios to a given node) adds more capacity to the mesh. The only good counterargument here is that more radios obviously cost more money, but this suggests that successful mesh product lines will scale from single-radio nodes to those with multiple radios, allowing the network to growth with little or no replacement of those nodes already installed (although such would obviously be the solution if convenient places to mount the nodes are in short supply). The bottom line, though, is that meshes should be able to handle voice traffic with little problem.
I stand by my prediction that VoFi will become one of the major applications for WLANs, and a key driver of both the enterprise and outdoor/public access/hot spot markets overall. I'll have more on how this becomes reality later this year.