Tuesday, June 19, 2007

"Cheap wi-fi too slow"

"Cheap wi-fi too slow" so says :
Bill Tolpegin is vice-president of planning and development for the municipal networks unit of Earthlink, a US-based company that built municipal wi-fi networks for cities including New Orleans, Philadelphia and Anaheim and has been asked to devise plans for networks in San Francisco, Houston and Atlanta.
This is in line with Lafayette Pro Fiber's long-held position—wireless broadband as currently conceived is not a viable substitute for a wired network. But that's a pretty shocking comment coming from a major player in the muni wifi business who has been selling wifi as if it were a subsitute (not an addition) to a powerful wired system--What Tolpegin is saying is that his companies networks are too slow. Why? The answer is instructive for Lafayette:

He says the wireless mesh technology advanced as enabling wi-fi to quickly and cheaply cover wide areas can only do so at very slow speeds...

The mesh is slow because it relays data from access point to access point, he says. As traffic hops over these networks the available bandwidth is quickly consumed relaying data back onto a faster, wired network, greatly reducing the bandwidth available for each user.

The only way to get around this problem, Mr Tolpegin says, is to create "injection points" on mesh networks where data is transferred to a different network in order to relieve the wi-fi mesh of the need to carry all data, all of the time.

Translated: mesh networks need to consist of less mesh and a higher percentage of nodes that tie directly into the high-speed, wired, backbone. Mesh technologies, which promised a cheap infrastructure built on few--hence cheap--backbone connections isn't panning out in practice.

There's more:

Earthlink has struggled to find commercially viable ways to make the task easier. "Nobody has high-bandwidth, low-cost networks that deliver," he says. "They are not telling the truth, not even the WiMax vendors."

The answer, he says, is far denser deployment of wi-fi access points.

So Earthlink's hard-won experience tells people two things about building high-speed wireless networks:
  1. minimize the mesh, work as close to the wired backbone as is possible
  2. maximize the density of the nodes
Lafayette's unique situation—with the wifi provider running a massively capable fiber network down every street—allows us to take a slightly different perspective on these truths. Because we will make an extremely capable network available to every user at a very reasonable price there will be little pressure to make the wireless network in Lafayette struggle to provide "dsl" or "cable" equivalent capacities for fixed uses. We'll have fiber at our fingertips for in-home and business use. Should we want wireless inside our house we can easily provide it for ourselves. The wireless network can be "freed" to be the wireless, mobile extension of the full network—not a low-priced substitute for it.

With LUS both the fiber owner and the wifi provider, it relatively easy for Lafayette to follow Earthlink's advice about minimizing the mesh. Earthlink has to pay, every month for every drop off the hardwired network and for the bandwidth it consumes. Lafayette will only have to pay once for the hardware drop and the incremental cost of using that bandwidth will be very nearly zero to the extent that taffic remains within the LUS network . Earthlink and LUS will be in radically different fiscal postures and the advantage is all to LUS (and her customers). In fact, LUS already appears to be planning to minimize the mesh in its network—the intial order was for a 1:1 ratio between fiber fed nodes and "radio-only" nodes. (The story says that Earthlink is struggling toward a 2:1 ratio between backhaul feed nodes and radio-only ones.)

Earthlink's advice about node density is, no doubt, also a good one. I've no idea how densely LUS is planning to pack our network. But it is worth noting that what they are buying with denser placement is faster speeds--wifi speeds fall off dramatically as you move away from the node. Because Lafayette's wireless mobility system will not be burdened with being an adequate subtitute for a DSL or cable system—as it is when it is introduced as a cheap alternative to those products—we'll be able to consider a density that works best for wireless' unique mobility functions. Currently those applications center around data and voice and require less bandwidth than video. (Though video, albeit small video, appears to be coming.)

All in all Lafayette's decision to emphasize building a fiber network as the best choice for a community network seems more prescient every day. Wireless is not an ideal technology for your primary network; its best role is to be hung off an advanced wireline network to serve those mobile purposes fixed wireline connections cannot fill. And, as an additional, ironic benefit it turns out that the most economically sustainable way to get a cheap, truly high-bandwidth wifi network is to commit to building your own fiber to the home network first.

Lafayette is doing it right.

No comments: