Pacific Connection(英語)

Mesh Wi-Fi Creating Municipal Network Gold Rush

Wi-Fi hotspots-those puddles and pools of Internet connectivity that cover coffee houses, hotels and college campuses-are poised to becoming large enough to cover an entire city. Municipal networks (sometimes called "metropolitan area networks") based on Wi-Fi technology are being seen as a new kind of public utility, like water and electricity, that would provide affordable broadband service to an entire community. Here in the United States, these networks are largely in the proposal stage, with companies competing to build and operate them. And while a few smaller networks have already been completed, many have been criticized for having disappointing coverage. Even so, many American municipalities are now looking into the idea and optimists predict that within ten years, every American city of any size will offer wireless network service that reaches most of its residents. If these predictions are right, Internet connectivity will be a part of the air like television signals and oxygen.

The secret sauce behind most of plans for municipal networks is a technology called mesh Wi-Fi-in which data from multiple radio transceivers, typically perched on telephone poles, is sent to a wired, high-speed Internet connection. The idea of building mesh networks does not garner universal praise, especially from telco and cable companies who have already invested in delivering wired broadband connections. Last April, Ivan Seidenberg, chairman and CEO of Verizon Communications, told the San Francisco Chronicle that the city's plans to build a municipal network "could be one of the dumbest ideas I've ever heard. It sounds like a good thing, but the trouble is someone will have to design it, someone will have to upgrade it, someone will have to maintain it and someone will have to run it." Verizon is not exactly an unbiased source here. The telephone company offers DSL service for $23 a monthly for speeds of up to 3 Mbps or $18 for up to 768 Kbps. Here in the San Francisco Bay Area, cable company Comcast charges $43 a month for speeds up to 6 Mbps. (Actual speeds can vary greatly. As I write this early on a Friday morning, my connection speed is 2.8 Mbps.) Hence companies that once had the broadband market to themselves are looking at new competition.

This debate is also playing out in New Orleans. After Katrina hit, Greg Meffert, the mayor's chief information officer, patched together a wireless Internet connection using a single connection in the mayor's temporary office, based in a downtown hotel. "Throughout the Central Business District, wireless Internet service sprang to life incongruously amid the staggering desperation," wrote Kathryn Jezer-Mortoni in a New Orleans CityBusiness report. A VOIP connection allowed the mayor to phone out at a time when telephone towers, both landline and cellular, were down. Now, the city would like to maintain and expand that service, perhaps through an investment with Earthlink. But at this writing, the local telephone company is complaining of unfair competition, citing its own multi-million dollar investment in repairing the network as well as a local law that restricts municipal-owned networks to a piddling 128 Kbps.

Proponents of municipal networks argue that this technology is not so much competition as an affordable alternative to the entrenched broadband providers, intended mostly for the millions of people who are still using dialup. Various pricing schemes have been proposed, including free service subsidized by advertising. In San Francisco and Philadelphia, Earthlink has proposed $20 a month for speeds of up to 1 Mbps. That would be a good deal for many people, including existing broadband customers looking to save money. The selling proposition is a relatively cheap broadband connection that is fast enough for most websites. And of course Earthlink itself is bidding on the idea that it, not the cities, would be responsible for building, maintaining and running the service-either directly or through partners. Another company, MetroFi, is proposing the same for Portland, Oregon and some of the cities that comprise California's Silicon Valley.

What this comes down to is a tradeoff between the companies who invested in a broadband infrastructure based on an expected return, and other companies, using a newer technology, who think they can provide nearly comparable service for less. But fair or unfair, mesh Wi-Fi proposals are popping up everywhere because broadband Internet is no longer seen as a high-priced frill, but a near-necessity-as fundamental to the American way of life as television. (Whether either is truly necessary to a fulfilled life is another matter. I know a few monks in India who have other ideas.)

Columnist Glenn Fleishman, who writes the Wi-Fi Net News blog (, traces back the idea of a true city-wide network to August 2004, when Philadelphia mayor John Street announced plans to build for a city-wide Wi-Fi hotspot. Philadelphia was a good candidate. In a world that is increasingly going broadband, the city still had 72 percent of its Internet-connected households using dialup and a substantial number still have no household Internet connection whatsoever.

While actual, operating municipal networks are still hard to find, a few smaller cities have come online. Fleishman cites at least one success: Tempe, Arizona. With its mostly flat, desert expanse, Tempe was born to be a hotspot. The city started with equipment placed on buttes on both sides of the town to serve its government buildings, thereby avoiding the cost of high-speed digital lines from the local telephone carrier, Quest. Working with a local networking company, the city then added a mesh network with about 650 access points to extend broadband services to residents. The completed network will reach 90-95 percent of its service area and is slated to cover 187 square miles by the end of the year. The city now claims to be the first citywide hotspot in the United States.

But Fleishman thinks that Tempe's success does not necessarily mean that other cities will do as well. "Using mesh networks to connect up a flat, sprawling, relatively sparsely populated suburb is one thing," he wrote in The Economist. "But many unknowns remain, given the scale and nature of the networks about to be installed in Philadelphia, San Francisco and several other large American cities-all of which will push Wi-Fi technology to its limits." That has been the case in Charleston, North Carolina, where coverage so far has been less than expected.

Even smaller hotspots are also serving populations that were previously too small to attract any kind of broadband service. Chesterhill, Ohio, population 312, now has broadband thanks to an antenna mounted on the village's water tower. The town of Scottsburg, Indiana, population 6,040, offers its own brand of broadband service for $35 a month after failing to attract Verizon. Fourteen Indian reservations near San Diego, California are now served by a network that uses solar-power antennas-at network speeds many city dwellers would envy.

Fixing the holes in the mesh

The tipping point for mesh Wi-Fi has come with the dropping cost and rising performance of the mesh radios or ("routers," depending on who is selling the equipment). "That has contributed to better coverage-because you need a certain density to get coverage indoors-we think about 30 to 36 nodes per square mile," said Cole Reinwand, Earthlink's vice president of product strategy and marketing for Earthlink Municipal Networks. "The lower cost of the radios makes that economically viable. If you are trying to sell service at $20 a month, and you factor in the capital expense in putting in 36 nodes per square mile, the cost per node has to be low enough that you can recover your investment with reasonable signal penetration rates."

Reinwand is in a good position to argue the point. Earthlink is the front-runner to sponsor the two most highly visible citywide networks in the U.S.: Philadelphia and San Francisco. The company is also involved with projects in the California cities of Anaheim and Milpitas, and New Orleans, with proposals submitted to Long Beach, Pasadena, and Arlington County, Virginia. The company is building a small pilot project in Minneapolis.

Earthlink has primarily been known as an ISP, offering both DSL and dialup-but its involvement with the Philadelphia network project put it on the map for municipal networks. "When cities first considered municipal networks, they looked at funding it themselves, hiring a subcontractor to build and operate the network, then wholesale out access to ISPs," Reinwand said. "In Philadelphia, Earthlink changed the model in a drastic way-and now it looks like other cities are following this model. Philadelphia was the first major city to go with a public/private partnership. Under the offering, Earthlink would actually fund the network, rather than the city. In exchange, we would own the network while providing revenue-sharing fees to the city. Some of those fees would be used to subsidize accounts. In Philadelphia, for example, Earthlink will provide 25,000 accounts available for $10 a month."

Aside from funding the project, Earthlink is also involved with network design. The company has a partnership with Motorola, which would be the general contractor and provide some of the equipment. Reinwand describes the Earthlink architecture as a "hybrid mesh: a point to multi-point network." It begins with a set of transceivers that are typically mounted on city telephone poles or light posts. To upload the data, a Motorola Canopy gateway is paired with a transceiver, transmitting the aggregated data to access towers that, in turn, transmit the data to Earthlink's national backbone, which has 18 major points of presence and 150 minor "POPs" around the U.S. Two different methods are available for this data "backhaul": a terrestrial fiber circuit and high bandwidth, leased microwave.

This mesh scheme also allows for a wireless T1 alternative for schools and city buildings. Reinwand said the cost would be about 25% to 50% of a normal landline T1, with much faster provisioning: typically less than seven days versus the usual 60-90 days. "Our advantage is that the network already is in place. All we need to do is attach a radio to the side of the building." Earthlink's San Francisco proposal has an additional twist: Google is proposing a free 300 Kbps service, funded by advertising, with traffic consigned to a separate virtual network. The proposal has generated some controversy. Critics charge that the 300 Kbps speed is more appropriate for 1996 than 2006, and that the entire network could interfere with other freely available wireless networks.

I asked Reinwand about sporadic press reports that mesh networks don't always deliver on their promises. In Charleston, South Carolina, for example, a $50,000 network scheduled for rollout at the end of 2005 is only reaching 20 to 30 percent of the residents. Others may have to buy signal-boosting antennas in order to pick up the signal indoors. And in Chaska, Minnesota, the is billed as "Always on, Anytime, Everywhere in Chaska," at least according to the homepage. But when the service debuted, about a quarter of the town's residents were getting substandard service. "We do see some variation in speed among customers," said City Manager Dave Pokorney in a CNET interview. "And that has been something we've struggled with from the beginning. At one time we were seeing people getting signals of 250Kbps, and someone six houses down was getting 1.2Mbps. That just wasn't acceptable." He said that an antenna upgrade has dramatically improved service.

Reinwand said that the problem with some of the early networks is in underestimating the density of the mesh. "You'll see companies low-balling on their offer to the city. But they are failing to mention that they are not building a network that is capable of providing indoor access. They have to control their costs, and they do that by supplying fewer radios."

Reinwand cited St. Cloud, Florida, as another example. "That's a city-owned and built network, but they had only deployed about 17 nodes per square mile and a lot of the residents were complaining about spotty coverage and inability to connect indoors. Those are exactly the types of symptoms we would expect from a network that wasn't deployed with enough radio density." Nevertheless, the city expects a $650,000 annual savings for government communications on its $2.6 million investment, in part by providing wireless voice services to its employees. Howard DeYoung, St. Cloud's director of information technology, told the New York Times: "We're going to be doing things with public safety, and mounting video cameras in parks so our police force won't have to be everywhere in order to keep the city safe."

Reinwand said that another problem with early networks is that they use an all-Wi-Fi mesh topology-in which each node of the mesh not only communicates with users, but also is responsible for carrying aggregated data upstream to the gateway. Hence the nodes closest to the gateway become a data bottleneck. Moreover, downstream nodes also share "talk time" with other nodes in the mesh path. That means that one node must pause and wait for another while it finishes its communication. Earthlink's approach is to place a Motorola Canopy backhaul gateway for every few Tropos Wi-Fi nodes, thereby reducing the number of hops before it enters the backhaul. The network also routes traffic from the light poles up to towers and rooftops, away from foliage and other obstructions. "Typical mesh Wi-Fi deployments attempt to use 5.8 GHz backhaul at street level, where it very often runs into significant levels of attenuation and interference," Reinwand said.

San Francisco poses special problems for a mesh Wi-Fi topology. The 49-square mile city has neighborhoods like Telegraph Hill and Russian Hill, which are built on terrain so steep that some of the sidewalks are actually stairways. Outside of the canyons of Manhattan, if a mesh Wi-Fi network can make it here, it can probably make it anywhere, and Earthlink is not arguing that it will be easy. But the company says that a planned 1,700 radio nodes, each of which can send signals 600 to 1,000 feet, will fill in the gaps.

As for the pricing structure, Reinwand said that the $20 monthly fee proposed in some cities is calculated to attracted dialup users, with the hope of upgrading some of them to higher speeds later on. "The 1 Mbps speed is not the fastest available, but it is certainly enough to provide an excellent broadband experience." He thinks that the service will be complementary to cable and DSL, rather than competition. "In fact, we've been talking to the cable and the DSL telcos about bundling these Wi-Fi services with their existing offerings-to allow users to have a high speed connection at home and take their connection with them when they go out around the city. We're open to that.

"Another thing that makes Earthlink's proposals quite unique is that we're committed to an open access business model: having competitors operate over the network. After building out the network it's going to behoove us to put more users on the network to help us recover our investment. We want to avoid building an airplane, only to fly it half-empty. Multiple providers selling services and filling up the capacity of the network will help us monetize it faster."

I asked Reinwand what lessons from America might be exported overseas. "The tricky issue for us is the different requirements around the world for the use of the Wi-Fi. In Europe, for example, the power output allowed by a Wi-Fi node is much, much less than it is in the U.S. This would require you to deploy significantly more radios per square kilometer, by a factor of 10.

On the other hand, Nortel is deploying a massive mesh Wi-Fi network in Taiwan that is scheduled to reach 90 percent of Taipei's 272 square kilometers with some 10,000 access points. The company claimed it had reached 50 percent coverage last year. Nortel says that ultimately, the M-Taiwan project will provide networking to 10 Taiwan cities, although it is unclear whether the some of these will primarily serve government agencies. Nortel equipment is also being used in Moscow, where 5,000 access points will serve some 3.9 million households. Tests are scheduled to be completed this summer. "If the news of a network of this scale being completed so fast with so many mesh access points came from a company other than Nortel or one of their peers, I'd be checking to see whether it was a hoax," wrote Glenn Fleishmann, on his blog.

On the horizon: WiMAX

Wi-Fi mesh is catching on in the United States primarily because it is so practical-Wi-Fi cards are now found everywhere. But some argue that cities should wait for a newer Wi-Fi standard, 802.11n, or for WiMax, an emerging technology based on IEEE 802.16 and ETSI HiperMAN wireless standards. When it debuts, WiMAX will be a wireless point-to-multipoint broadband technology. Unlike Wi-Fi, it is designed expressly for long distances and Motorola is already taking a WiMAX-like approach for its backhaul connection.

"Today, if you go to a rural town or suburb with no fiber optics or existing broadband, WiMAX, not city-wide Wi-Fi, is the better choice," said Glen Fleishmann. "You stick a WiMAX baystation on top of the water tower connected by fiber optics, and people pay $200 to get the equipment installed on the roof. " This scheme will deliver several megabits per second. One of the best-known providers is the Israeli company Alvarion, which launched its WiMAX-ready platform, BreezeMAX, in June 2004, after acquiring InterWAVE, a leading provider in the U.S. The other market, said Fleishman, is a less costly alternative for wired T1. "The plan in Philadelphia is to sell fixed WiMAX both to the city and businesses, running at T1 and higher speeds, replacing something like 700 leased wire line connections."

Fleishman says that the U.S. market is notably behind Japan, which has much higher DSL performance, with more coverage and lower cost than in the U.S. "And their 3G cellular networks are much further along, giving them a good sense of what mobile, high-speed Internet would be like." But the real action may be happening in South Korea, which is in the process of deploying a mobile version of WiMAX, called Wibro, "that will give them high-speed, robust coverage that is intended for metropolitan coverage." In a trial last May, testers found they could get broadband connections while sitting on a bus. A BBC reporter in Seoul wondered why a already nation wired for speeds between 20 and 50 Mbps, with 3G cellular phones, as well, would want wireless broadband. "You'll be able to watch TV on the move, while e-mailing and downloading big files at the same time," said a senior manager at KT, the telecommunications company behind Wibro.

Of course: wireless television. How did we get through a day without it?