Receive new posts as email.
This site operates as an independent editorial operation. Advertising, sponsorships, and other non-editorial materials represent the opinions and messages of their respective origins, and not of the site operator.
Entire site and all contents except otherwise noted © Copyright 2001-2009 by Glenn Fleishman. Some images ©2006 Jupiterimages Corporation. All rights reserved. Please contact us for reprint rights. Linking is, of course, free and encouraged.
Meraki Networks raises $20m in additional funds, spends a small portion on providing free Wi-Fi across San Francisco: Meraki, founded by MIT students, funded by Google, Sequoia Capital, and others, will expand its current 2 sq mi mesh Wi-Fi network in San Francisco across the whole city. Meraki will foot the backhaul bill and pay for equipment--but it won't pay for real estate. The city will help publicize the network, but I haven't read anything concrete about the city's plans. (Read good local coverage in the San Francisco Chronicle.)
The clever bit here, and how Meraki may succeed where EarthLink failed, is that the firm is relying on individuals and businesses to choose to opt in, site equipment, and take advantage of the network working better for everyone because they participate (paging Ayn Rand). Meraki plans to offer solar-powered outdoor nodes for extending the network's reach, which means potential locations don't have to provide electricity on rooftops or elsewhere, which in turn means fewer or no layers of approval from anyone in authority (whether a neighborhood association, landlord, or the city). "It's relatively easy to install on private rooftops," said Meraki CEO and co-founder Sanjit Biswas in a briefing earlier today.
In the process, Meraki gets a city-wide testbed for local search, local ads, and new technology. "The great thing about having a real-world testbed is you can see the performance," Biswas said. Meraki is testing advertising now, but Biswas says it's "very much in the test mode."
While Biswas didn't disclose any of the costs of building the city-wide San Francisco network or its recurring bandwidth bill for backhaul--he told the AP "a few million"--the math works for Meraki. Giving away 10,000 to 15,000 nodes that cost them as little as $25 assembled for indoor nodes and a few hundred for outdoor nodes with solar chargers and batteries doesn't add up to a lot. They don't have to negotiate pole rates, handle installation, or pay recurring venue fees. Bandwidth is relatively cheap if you can choose the points at which you inject it, which Meraki will be able to do. Update: I misunderstood the backbone part. Meraki will, in fact, install some hundreds of solar-powered outdoor nodes to run the backhaul; but they'll still be working with people to find those locations rather than securing those rooftops themselves.
Meraki has learned quite a lot about the real world, Biswas said, with their current San Francisco network, having identified 20,000 interferers--other devices within range and frequency of Meraki nodes--within the 2 sq mi area. Meraki nodes use a centralized intelligence to control routing, as well as a modification to Wi-Fi that doesn't affect end users' ability to connect, but does allow clusters of routers to act dynamically as a unit.
With unmodified mesh Wi-Fi networks, all devices within range of one another and on the same channel act in concert, reducing flexibility and throughput; smaller clusters produce better results. Further, most of the metro-scale Wi-Fi devices sold by Tropos and others are designed to put out the highest possible power output; Meraki uses generally low-power equipment. It's ants versus elephants, with ants being able to change course a bit more quickly. "It's kind of a brains over brawns approach. It's really because we have so many radios," said Biswas, that their network is more flexible. "We can set routes and load balance appropriately to get the maximum performance out of the network," Biswas said.
The current San Francisco network has seen 40,000 unique devices so far--there's no registration, so each unique adapter number is counted--and moved over 10 terabytes of data. Of those 40,000 devices, 1,000 were iPhones.
Biswas said that Meraki has discovered its biggest market may be developing countries where there's an established user base for the Internet that's limited to dial up or slow-speed broadband, and where carriers could deploy Meraki gear to get around non-existent copper infrastructure.
"A lot of our largest customers are carriers that are entering markets in Brazil and India," he said. "We're not at this point going after the most remote villages in the world; there are some very dense populations who...would love to use broadband but just can't get it."
The storm broke over Meraki Networks change in pricing and services over the weekend, but skies seem to be clearing a bit: I wrote yesterday about how the mesh networking firm was raising its prices for its node for networks that required user logins and wanted to charge users for the service, and how many community networks were up in arms because of a lack of notification. I was surprised by it all because Meraki has had a strong community focus. According to Sanjit Biswas, the head and a founder, a fair amount of miscommunication led to the ire expressed. They've spoken to many of the networkers involved, and things are calming down.
Biswas said this afternoon that although the firm had called about 850 customers before they pushed new material to the Web site on a tiered pricing structure and a transition plan for existing networks that they clearly missed some people who then felt that they were being told of a fait accompli. The forums exploded a bit, and Meraki didn't respond there initially. "It's really our fault for not posting more heavily on the forum," Biswas said. He's trying to remedy that now. The Web site soft launch was intended to promote discussion which Biswas said they would use to forge the final plan; it just came a little faster and more intensely than they expected.
The new pricing structure for Meraki leaves the $50 indoor and $100 outdoor nodes for Standard edition networks only, which have no fees for access, and have no user accounts. An ad bar is required. Biswas said that they didn't understand how much many of these networks rely on local sponsorship, often in-kind splash screen advertising, and they'll modify the the Standard edition to allow multiple splash screens, as well as sponsorship messages in the ad bar that will appear in Web pages. They're also adding back more extensive whitelisting, which allows a device without a Web browser to connect to a network without viewing a splash page.
The Pro edition, which features $150 indoor and $200 routers and includes six months of email support, had a mistake in its description, which was part of what angered many forum posters. Biswas said that displaying advertising that Meraki collected revenue from would be an option not a requirement for Pro users. That takes a bit of the sting out there, I'd imagine. Biswas said that until they have a sense of volume and behavior, Meraki doesn't want to promise a revenue share as they don't know how much or in what form they can offer it.
This fix helps differentiate the Standard and Pro edition better, clearing up a point of confusion on my part: Standard is ad-subsidized; Pro is full support, full cost.
What Meraki calls Legacy networks, which use nodes sold before the changes go into effect, will have more options for how they proceed, too, instead of what was previously a choice to add new Pro nodes or convert the entire network to Pro. They're working with individual networks on these issues. "Existing customers who bought stuff from us don't have to change," Biswas said.
Biswas had earlier told me via email that part of the issue with raising price had to do with resellers, who they expect to rely on more in the future for sales. At $50, there wasn't enough money in the system for a retailer to package a system with support or discount nodes for volume. At the higher price, where Meraki expects integrators and resellers to include support, there's a lot more flexibility. Biswas said that he expects to see Meraki Pro edition nodes sold below $150 on a one-off basis by retail partners within a few months, as well as resold by network builders. Outside the U.S., he said, channel sales become particularly important. "There's something to be said about buying in Brazil from a Brazilian reseller," he said.
Biswas noted that in the future, they'd try to communicate changes in advance of those changes, so that their customers didn't view updates as the final word before that was the case. "Lesson learned," he said.
Meraki has changed its pricing and feature model for its mesh networking system, angering early users: Exiting its beta, Meraki has changed its pricing and service model, while requiring the display of advertising and a piece of the action for handling billing. This abrupt change, announced quietly last week, has resulted in a nascent networker revolt. It may be that early infrastructure builders abandon Meraki because to continue expanding networks, their cost structure has gone way up while control has gone way down.
The two basic nodes that Meraki sells, Indoor and Outdoor, were priced at $50 and $100 during the beta. The beta flag is gone from their site, and while you can still buy the nodes at that price, many features formerly included raise the price to $150 for Indoor and $200 for Outdoor. These prices move them into the range of what used to be much higher-priced metro-scale hardware.
Meraki's system works by having little intelligence built into nodes instead relying on an Internet-hosted administrative system to handle the heavy lifting. Adding nodes requires just some configuration via a Web site. Each node auto-discovers existing networks allowing their addition. Each gateway to the Internet added to a Meraki network increases the bandwidth pool.
The change puts tiers into effect. The Standard edition with the cheaper pricing doesn't allow billing nor user authentication, nor does it offer a custom splash page (a logo is all). Even the network name gets Meraki branding on it ("Free the Net" precedes a local name). Standard does offer a private WPA-protected network for the owners, but without any granular access control. This means that a group that wants to change small amounts for access or control access by user cannot use the entry-level version. According to the forums, this includes a lot of early networks. Standard networks also can't "whitelist" more than a handful of devices, which means that devices that lack the ability to display a Web page for authentication purposes can't even access the network. So no Xboxes, VoIP phones, and so on.
The Pro edition bypasses these limitations, essentially adding a $100 per node license for advanced control and the ability to bill. However, because several of the Pro features were previously in the only version Meraki sold during its beta, many networks had build their business model, their pitch to cities and communities and businesses, and their financing on $50 and $100 nodes. The Pro version ups the ante while taking away some of what these network builders need, according to the forums. (A Carrier class version, with no pricing disclosed, allows a tie-in with existing user authentication system, and allows more of a private-label version.)
So-called Legacy networks, ones that were built before this last week, can continue to run as they do or be upgraded to Pro status at no extra cost. But once upgraded, all the Pro limitations are in effect. Any new nodes added to a Legacy network must be Pro nodes, too; there's no option to add Standard nodes. This is part of what's causing ire on the Meraki forums. (One of these networks, Kokua Wireless, was praised in an editorial in today's Honolulu Star Tribune; someone from the network noted on Meraki's forums, "It'll be hard to build a business even @ the Pro level knowing Meraki can make such drastic changes." One network operator has already flashed his Merakis with alternate firmware.) Update: Meraki says it's been talking to all the operators and groups that have problems with their new model; I'll have more details Monday.
In this formal release, Meraki now displays advertising on every page using what they variously call Community Messaging, Community Messaging & Advertising Platform, and Messaging Platform. It's often euphemistic. In effect, Google ads--Google is a Meraki investor--will appear on every page in a special toolbar that requires no installation. Which means that it's inserted using some technique--probably frames--that might cause other problems, too. Network operators can insert localized messages, but to a limited extent. There's no option for Standard and Pro users to disable advertising or the messaging bar. There's no disclosure on what revenue, if any, Meraki splits with the network operator. If Meraki handles billing for Pro users, they keep 20 percent of gross revenue.
I had to check the Web site after being alerted to this change, assuming that senior management--largely a bunch of MIT grad students who turned RoofNet into this clever commercial offering--had been fired or shuffled, and a new carrier-oriented CEO was hired. As far as the site shows, it's the same gang. Which surprises me, given how open the group was about communicating, and how interested in community building. This is a huge stumble, when the community tells me that they were unaware of the changes to come until the Web site started to have new information on it.
At $150 and $200 for managed nodes, Meraki has now lost some of its edge against companies like Tropos Networks, the most likely comparison. While Tropos, Cisco, BelAir, Motorola, Strix, and SkyPilot typically charge $2,000 to $5,000 for their nodes, with the more expensive ones including 5 GHz backhaul radios, those prices are heavily discounted, several sources have told me recently. Tropos nodes can cost EarthLink, its biggest customer, under $1,000, I have it from multiple parties outside the service provider.
While $1,000 is clearly five times $200 for a similar outdoor piece of gear, that belies several factors. Tropos is using enormously high-powered radios, capable of producing up to the maximum legal signal strength in 2.4 GHz. Tropos has a sophisticated back-end that allows integration with many popular management and authentication systems. Using Tropos or other vendors' network gear is substantially less trivial than building one from Meraki pieces, but with enormously larger coverage areas relative to cost now, Meraki has moved itself towards that equipment range without bringing the robustness needed.
The system and hardware has gotten rave reviews for the simplicity in building a network in bits and pieces. But the good will may have just flown out the window. I have just sent a query off to Meraki's PR firm, but it being Sunday morning, I don't expect to update this post until Monday.
Meraki's inexpensive mesh-routing nodes are a hit: About 15,000 users are connecting to their $50 nodes ($100 for outdoor units) in 25 countries during their testing phase. The company didn't note how many routers were shipped, but it's likely between 1,000 and 2,000 based on their descriptions of density. Meraki's devices cost a tiny fraction of what metro-scale mesh networking equipment costs, and that's partly because they have fewer features and reach much shorter distances, requiring denser installations. But the point isn't necessarily blanketing a city, but rather putting a signal over a neighborhood, a village, or an apartment building. Meraki is a bit like powerline networking: Covering connected areas without a lot of infrastructure. You can listen to an archived podcast interview I conducted in Oct. 2006 with Sanjit Biswas and Hans Robertson, two of the co-founders.
Meraki received very positive coverage in Randall Stross's Digital Domain column yesterday in the New York Times, in which he looks at how the next billion people will be connected to the Internet. Stross compares the efforts of city-wide networks to outdated methods of lighting cities by using huge arc-lights high in the air. He cites a practical test in Portland, Ore., where 400 apartments were served by 100 Meraki routers, which works out to $13 a household for installation, and, Stross writes, about $1 per month per household for Internet access. Because this rollout was in association with a non-profit, that group is obviously dealing with the labor costs of installation, tech support, and maintenance. (Even with a very low failure rate, 100 routers might see a 1% failure per month or higher, especially in high-use conditions.)
Finally, Meraki announced $5m in first-round funding led by Sequoia; other investors weren't noted. GigaOm mentions that Google and Sequoia invested in Fon; Sequoia in Ruckus Wireless; and Benchmark Capital in Whisher.
Two of the wave of newer metro-scale wireless equipment firms release news on deployments: Wavion, InspiAir, Go Networks, and Cohda Wireless are in the wave of startup vendors that came in being after BelAir, SkyPilot, Strix, and Tropos became established (but still startup) players. (Cisco and Motorola may have acquired and built metro-scale and mesh technology, but they're hardly startups.) These firms all employ techniques to allow greater range and greater spatial reuse of frequencies than the established players, typically via multiple antennas to allow beamforming and/or spatial multiplexing (unique data streams that follow disparate spatial paths using the same frequencies).
Wavion notes today that CONNX, a Maryland network operator, will deploy 120 of their APs. Their AllCoNet (Alleghany County Network)is one of the earliest municipal wireless networks, and the organization did rigorous testing of a host of mesh and metro-scale devices. The group claims that just 12 Wavion devices (with six transceivers and six antennas) cover a square mile. In most competing deployments, 25 to 50 APs are recommended, according to my many sources on this matter.
Wavion has said in the past that they plan to add SDMA, a method of beamforming a signal to specific clients in such a way that other clients on the same frequencies receive nearly zero energy, enabling simultaneous reuse of frequencies over space. That SDMA mode isn't yet out, and some engineers I've spoken to think that the current Wi-Fi MAC can't accommodate this mode. Wavion has acknowledged it's a challenge, but they believe it's got a solution.
Meanwhile, InspiAir, a company that boasts in its press release that it violates people's notions of the laws of physics, has formally unveiled a network covering Helsinki, Finland's central core. The network was biult with local firm Omni-Directional Communication Products (OCP). The firm says they've monkeyed with standard Wi-Fi to support dramatically higher areas of coverage with good throughput.
The press announcement says only 14 access points were required to cover the area in question, the size of which isn't defined, and thus that number provides no guidance in comparison to other systems. Update: Esme Vos has the details: It's a 4 sq. km (1.5 sq. mi) installation, but they told her only 2 APs are being used; their press release distinctly says 14. They also told her indoor coverage was hit or miss. If you're not attempting full indoor coverage, their competitors would also recommend a lower range of APs per square mile.
They note their APs use 60-degree sectorized antennas, offering 1.6 km reach in near-line-of-sight (nLOS) positioning. Their network operates at high 802.11b speeds--they suggest 6 Mbps--with fast 5 to 7 millisecond handoffs for mobile users. Service will run about €2 ($2.70) per day and €15 ($20) per month.
Cisco combines hardware, design into tested packages for metro-scale deployment: The company announced their plan today to sell integrated packages of their wireless mesh equipment, wireless management systems, and routers that would let metro-scale service providers avoid building each deployment from the ground up each time. Joel Vincent of Cisco said in a briefing last week that this streamlined approach--known as ServiceMesh--will allow service providers to go from answering a city's request for proposal (RFP) to "having revenue as soon as possible."
Cisco has optimized their package to integrate applications across the network, including such popular municipal examples as wireless meter reading, in-the-field building inspection reports, and general public safety communications. The idea is that with Cisco providing help on the glue that binds these top-level applications from a variety of existing municipal and enterprise vendors into the wireless network, service providers can mix and match solutions without reinventing the wheel. This would result in more consistent deployments at lower cost.
Cisco's own services division can provide the design help necessary to put the integrated package together; Cisco's partners, like IBM, will also be able to bundle and resell this new offering.
This new offering is tied with Cisco's announcement that eight medium-sized cities have been deployed through service provider partners with packages derived from this reference design. Vincent said that most of the upcoming city RFPs would be from sub-500,000-population municipalities where there's less of a user base across which to spread design costs, making efficient planning and upfront integration more crucial. The international market is also heating up, and Vincent expected that a fair amount of future business would come from outside the U.S., citing Singapore's near-term effort to unwire the entire country as one example.
Vincent likened ServiceMesh to the way in which hardware products move from chip vendors into a broader marketplace. "If you think of a chip company that wants to spark the Wi-Fi revolution, the first product to come out was probably a reference design from the chip company. Then all the product companies took it and productized it," he said, turning it into an item to sell. When a service provider asks, how do I add wireless Internet access to a city, Cisco can provide that reference design as a well-thought-out starting point with the majority of pieces in place, Vincent said.
Vincent made the fascinating prediction that most networks on this scale would wind up offering Internet access at no cost, with the area that service providers derived revenue from coming solely from specific municipal, business, and consumer applications that would run over the network. "The days of charging money just to access the network are slowly disappearing," Vincent said. "Service providers have to stand back and say as the cost of pure access and [voice] minutes decline," where will revenue come from.
Cisco has worked closely with IBM to combine the design and deployment of networks, with the two companies finding a neat fit between Cisco's architecture and IBM's application focus. The two are part of two MetroConnect consortiums: the first, with Azulstar and Seakay won the Wireless Silicon Valley bid; the second, with Azulstar gave them Winston-Salem. In the former case, over 40 separate municipal entities, mostly cities, will strike separate deals with the consortium, and should give Cisco's approach an excellent test bed for both ease of deployment and adding other features to the mix.
Interesting development in the cost structure for mesh-based architecture: The folks at Champaign-Urbana Community Wireless Network (CUWiN) have ported their open-source mesh routing system to the Meraki Mini platform. The port allows the CUWiN software to operate on inexpensive nodes. Meraki hasn't set its release price for its gear, but it's about $50 during beta testing for those to whom it's supplying equipment. This is a nice convergence, as Meraki's codebase is derived from MIT Roofnet, where two of the company's founders come from as well. MIT Roofnet and CUWiN were working in parallel towards similar aims.
I interviewed two of Meraki's founders in a podcast last week, in which we talked, in part, about the commoditization of node equipment. For instance, RoamAD down in New Zealand transformed its business a bit ago into selling its software and management platform; its code runs on commodity Wi-Fi gear that can cost 1/10 to 1/100 of metro-scale vendors who sell complete hardware and software solutions integrated together.
Metro-scale networks are all the range, but Meraki aims small: The company, emerging out of the academic project MIT Roofnet, has mesh-routing algorithms that work on commodity devices, and they aim to target markets like apartment buildings and hotzones. In this podcast, I talk to two of Meraki's co-founders, Sanjit Biswas (president) and Hans Robertson (chief operating office), about how mesh networks work, MIT Roofnet's algorithms, and who might want to deploy small-to-medium mesh networks.
In the podcast, they refer to an installation in Stapleton, Colo., using their equipment that was featured in a local TV report, archived on YouTube. [MP3, 17 MB, 35 min.]
Interesting column by UK Wi-Fi gadfly--say that three times fast--Guy Kewney, pointing out a competitive technology to Tropos's new approach: Kewney thinks I was too broad in stating that all Tropos competitors choose to use 5 GHz for backhaul only, where Tropos's new metro-scale mesh system will use 5 GHz as an alternative mesh routing path with dual-radio nodes. He writes that LocustWorld, in operation since 2002, offers similar mesh-over-multiple-radio technologies. And that the firm is ignored because they don't really sell hardware or software; they develop software for reference platforms and sell consulting services.
Because LocustWorld doesn't have salespeople or an obvious US presence in the market--because they're not per se selling a product--I haven't seen any comparisons with their software running on commodity devices against commercial, expensive gear sold by Tropos, BelAir, and others. I don't know whether their approach has been considered by any of the major operators now in the US market. MobilePro has chosen to use Strix gear, EarthLink picked Tropos and Motorola, and MetroFi has stuck with SkyPilot.
It's worth noting that both the commodity devices and the multi-thousand-dollar, multi-radio proprietary hardware mostly use Atheros chips. That's right. The heart is the same. Tropos's VP of engineering Saar Gillai said in an interview earlier in the week that Tropos has put significant effort into building their own radio boards around Atheros chips, while, he asserted, their competitors typically purchase prefab reference design-based boards that Gillai says vary in performance from board to board.
If you look at Defacto Wireless, a partner of LocustWorld that sells Atheros-based mesh nodes in the US that will run LocustWorld's software, they offer single-radio ruggedized external nodes for $700 to $800, which is about a fifth of the list price of a Tropos single-radio node, which Inc. magazine said retails for $3,500.
RoamAD from New Zealand also offers software for fungible Wi-Fi boxes, although they charge license fees for it. Martyn Levy, RoamAD's CEO, told me recently that their mesh approach allows their value-added integrators to choose the best set of Atheros-based hardware, manufacture or purchase it in whatever quantities they want to negotiate, and then install RoamAD's routing gear.
I'll be curious if someone performs the testing necessary to see how RoamAD and LocustWorld stand up to their much-more expensive, proprietary hardware brethren. I have been told by RoamAD and LocustWorld seems to maintain that a large number of nodes using their respective software have been installed worldwide.
But no significant metro-scale operation in the US has been built on either platform, and we're living in the crucible of municipal Wi-Fi right now. RoamAD has a large installation in downtown Perth that's focused on three-dimensional (i.e., tall building) mesh networking, but it's only 20 blocks. Bigger networks are in the works, but not yet announced.
It's no surprise that Tropos would offer multiple-radio mesh nodes for its metro-scale networking equipment, but how they're using the second radio is unusual: It has been abundantly clear for many months that for Tropos Networks to compete effectively for metro-scale network deployments they would need to add models of mesh network routers that had two or more radios. Tropos has now done so, with a model slated to ship in October that contains both a 2.4 GHz and a 5.8 GHz radio. What surprised me is how they chose to use that second radio, and how relatively little they're charging for it. (This announcement was slated for Thursday, and News.com broke the embargo this evening.)
Tropos has chosen to upgrade its MetroMesh operating system to support mesh routing across bands. The company said in a wide-ranging interview that their new approach will choose the most efficient route to move packets among clusters of mesh nodes. In an extreme example, if the user-facing 2.4 GHz Wi-Fi network is busy, then Tropos routers would move all intra-nodal traffic to 5 GHz hops; more typically, however, any given packet will transit over the least loaded connection, which could be hopping among a combination of 2.4 GHz and 5 GHz links.
Combined with a recent improvement that allows per-packet power level controls, this change could produce a substantially more efficient use of spectrum; lack of efficiency has been one of the biggest criticism of single-radio mesh networks in general and Tropos's system specifically. True mesh networks require all nodes in a cluster to use the same frequencies, which results in a single packet occupying network time slots usually at least once, but possibly several times, to reach a connection out of the cluster. With a second radio, packets could transit the mesh-only 5 GHz connections as need be, offering more time slots to clients. (In future software releases, Tropos might support clients on 5 GHz as well as mesh traffic.)
Their first two-radio router doesn't solve the problem of backhaul, and will still require an capacity injection connection to equipment like that from Alvarion or Motorola, which Tropos points out is typically substantially less expensive than a mesh node.
Saar Gillai, vice president of engineering, said Tropos approached the multi-radio problem differently than their competitors, because there's a high cost in "closing" 5 GHz links--or providing full coverage in every part of a network. Tropos has opted for omnidirectional 5 GHz antennas, which dramatically reduces a 5 GHz radio's potential range, but fits in neatly with Tropos' desire to distribute traffic across a cluster of nodes. In some clusters, 5 GHz radios will only cover 50 to 80 percent of the same area that 2.4 GHz radios cover.
This approach allows less dense areas to use single-radio routers and more densely deployed areas to employ dual-radio routers, rather than requiring every device have at least two radios. (Clusters could number from about three to eight units in which the 2.4 GHz and 5.8 GHz radios are using the same channels.) Tropos says this helps them maintain a cost advantage over competitors.
Tropos says that they can increase capacity 50 percent above their current levels with the same density of access points as they currently recommend for single-radio deployment. This gives them 50-percent closure, as they describe it, meaning that their 5 GHz links will only cover about 50 percent of the area. For a 20 to 30 percent higher cost, they suggest they could reach 80-percent closure, and nearly twice the capacity of their existing gear.
Bert Williams, senior director of marketing, said that deployment costs would be less because each 5 GHz radio wouldn't require alignment with other radios, as in point-to-multipoint systems, and adding capacity could involve simply upgrading their single-radio unit to a dual-radio model rather than building a new backhaul topology.
Not above some mild trash talking, Gillai stressed Tropos's development of their own radio technology, using industry-standard Atheros chips, but building their own radio boards and systems which reduced cost and allowed them to create what they characterize as more consistent performance from their radios. "Some of [our competitors] are using directional [antennas] because they don't have such a great radio," Gillai said. "If they're buying a whole radio in a reference design it's going to be more expensive for them. We invest much more in software." (I have heard this as well from wireless experts outside of the metro-scale networking business.)
The first hardware version of this technology in Tropos routers appears in model 5320, which supplements the existing single-radio 5210. Gillai said, "This is completely transparent to the users. You can add these routers to an existing network, or you can add these routers to a new network" as existing single-radio routers can receive software upgrades. He also said that existing networks under construction using Tropos gear would not see a shift in orders-in-progress, and that their partners had been aware for some time of the features and timetable of the new routers. Newer routers could replace existing ones in areas of congestion, although Tropos expects that busy areas would involve a denser deployment and dual-radio routers.
Tropos competitors--all of them--use a second band primarily as capacity injection, or the layer at which traffic from the user-facing radio in 2.4 GHz is offloaded to aggregated backhaul through point-to-multipoint connections. Now BelAir will say that they use intelligent switching in 5 GHz when they have two or more 5 GHz radios in a single box, and that's just fine. Strix has similar statements. SkyPilot uses extremely high-gain antennas to create point-to-point limits that are isolated in frequency and time from adjacent networks.
In their competitors' models, this allows substantially greater density at lower power of Wi-Fi nodes on nonoverlapping channels to provide a lot more bandwidth per square mile; a true mesh network like Tropos's requires all radios on the same channels, and thus you can only shrink or grow the area of a cluster of nodes (or reduce the number of nodes in a cluster) to improve bandwidth. However, the cost is prohibitive for high densities in any of the currently envisioned metro-scale projects using non-Tropos gear because each competing radio includes at least one 5 GHz radio for backhaul--what Tropos has described as the "5 GHz tax." This makes it difficult for competitors' advantage in a honeycombed Wi-Fi deployment to be expressed in the marketplace.
What Tropos is saying with their multi-radio approach is that 2.4 GHz is typically not fully loaded at any given time and that 5 GHz isn't just a directionally appropriate band. Rather, they're trying to combine the best attributes of mesh networking--closely placed nodes with frequency reuse in which the best node responds with the strongest necessary signal to a client--with the advantages of routing across dissimilar networks. The real question is whether in the field this amounts to the performance advantage that they cite, and whether the increased cost of their new units is justified by that advantage.
Tropos stated that their new 5320 would cost about 30 to 40 percent above their single radio units' cost, or more than 30 percent less than their competitors' two-radio units. It is nearly impossible to obtain pricing for metro-scale equipment--Tropos declined to provide dollar figures--and bulk orders and special partnerships also skew the numbers. Tropos also claimed in a briefing that their competitors substantially undercut equipment prices or even give equipment away to have a lower bid than Tropos-based projects.
The 5320 will ship in October, and the company expects versions with other technology, such as 4.9 GHz public safety band radios, to ship within the year. Future devices could include their own backhaul radios with directional antennas, too, and the company is watching MIMO closely as a tool to improve coverage areas. They are also watching WiMax, both fixed and mobile, to see whether there's enough service appearing in the market to offer either a client radio for backhaul or a base station radio as part of a mesh.
A Champaign-Urbana-based group gets $500K from the National Science Foundation: The Champaign-Urbana Community Wireless Network (CuWIN) group in conjunction with the University of Urbana-Champaign (UIUC) will use the grant to continue to develop their open-source mesh routing protocol system over the three-year life of the grant. The group wants to produce software that creates self-configuring, self-healing mesh networks with advanced properties that would allow clusters of nodes in which some cannot hear all others can still function with a high degree of reliability.
Tropos releases a software update for its mesh-networking hardware designed to improve interaction with handheld devices: The company said in a briefing earlier this week that Adaptive Mesh Connectivity Engine (AMCE) tweaks the mesh nodes' approach to dealing with different end-user devices on a packet-by-packet basis without any changes in the client hardware or software. Each packet, Tropos said, can have a different power output level, and can be customized with specific timing and framing characteristics that work best for a given device. The software is available today to Tropos users with support contracts. No hardware changes are needed.
Ellen Kirk, Tropos's vice president of marketing, said quite accurately, "There is no such thing as a standard Wi-Fi client." (It's also true that not all 802.11a/b/g devices are certified Wi-Fi, and thus a "Wi-Fi" device might not be that at all--it might not conform to interoperability standards required by that mark, which makes a "standard Wi-Fi client" an even harder mark to hit.)
The software also mitigates interference by analyzing the radio frequency noise floor and working around that to better "hear" incoming traffic and produce a usable outgoing signal of the right strength.
Later this year, Tropos will release an update that will use another bag of tricks--some of which I have heard of being deployed in enterprise-scale wireless LAN switched networks--that can force clients to associate with particular nodes rather than allowing the client its choice. While this is a hard-wired MAC (Media Access Control) issue that is very dependent on a particular implementation of 802.11 on the client, there are ways that access points can be clever enough to fool the adapter and make it stay put instead of hopping among multiple choices.
Earlier this year, Tropos released a set of extensions (TMCX) for customer premises equipment (CPE) device makers that would allow the CPE to talk to nodes and negotiate parameters and pass reporting and provisioning information. AMCE doesn't require CPE coordination to produce benefits, Tropos executives said, but rather reduces client variability on the network side by adapting to the Wi-Fi client's needs.
The HotZone Duo comes with one or two radios: The new products compete both against Tropos and BelAir's single-radio mesh nodes, and against BelAir, Cisco, SkyPilot, and Strix's two-plus radio products that offload backbone traffic through switched, mesh, or point-to-multipoint links. A few months ago, the Motomesh division head told me that residential use of metro-scale Wi-Fi couldn't compete against wired service, and was emphasizing the loss in single-radio hopping as packets were rebroadcast over clusters.
But, just as BelAir told me that they had to release a single-radio product to compete against Tropos for price and market niche, I imagine that Motorola wants to sell two-radio solutions that conform to their vision of metro-scale service.
These new products put them somewhat at odds with Tropos, which has been paired with Motorola's Canopy gear as backhaul for their clusters. In the Tropos configuration, a few mesh nodes, typically 4 to 6, are paired with a Canopy radio that uses WiMax point-to-multipoint aggregation to base stations that then in turn use license, high-frequency wireless to further aggregate to fiber points of presence and network operation centers.
Now, Motorola can sell an end-to-end solution, coupled with their existing multi-radio public-safety oriented system that uses both 2.4 GHz and the 4.9 GHz public safety band.
EarthLink contracted with Motorola as their service firm, meaning a division of Motorola handles integrating and maintaining all the technology for the metro-scale operations that will roll out in Anaheim (launching in a few days) and Philadelphia (test network due later this year). EarthLink committed months ago to Tropos and Motorola Canopy for its first five cities.
Tropos knew the writing was on the wall, which is why they now also have a partnership with Alvarion, one of the leading WiMax firms, with a history of broadband wireless gear dating back to the mid-90s.
Wavion unveiled its technology for using multiple-in, multiple-out (MIMO) antennas for city-wide architectures today: On the heels of Go Network's announcement at CTIA last month, Wavion's entrance could mark a trend. MIMO makes quite a lot of sense in cities in which dense radio frequency (RF) environments already occasionally make it difficult to operate a single new Wi-Fi access point, much less thousands.
Because MIMO offers greater range and greater discrimination through sorting out signals as they bounce across multiple paths from receiver to transmitter, this effectively--but theoretically until we can see benchmarks--provides more spectrum with less interference than competing technologies. In Go Network's case, they're using multipath to avoid interference; Wavion is focusing on spectrum reuse through beamforming, which focuses signal energy in particular places.
Wavion uses beamforming with six separate radios and antennas in their nodes; they haven't released full hardware specifications yet on their Web site beyond this, although they've developed their own chips. They also use SDMA (Space Division Multiple Access), which allows them to have effectively up to four separate unique beams operating at the same time for four separate users. That's downlink to users only, however; uplink would require proprietary equipment on the user's computer. This SDMA sounds an awful like like Vivato's beamforming system which used multiple radios and multiple beams versus multiple beams on a single channel, and I imagine we'll hear technical explanations as to how it differs in coming weeks and months.
Like Go Networks, Wavion isn't a mesh company. Go has some mesh capability, but Wavion connects directly to backhaul from its nodes. With fewer nodes, you don't need mesh for backhaul because you've already aggregated quite a lot of data. And metro-scale companies like Strix, SkyPilot, and BelAir aren't really using mesh for backhaul but a series of switched or dedicated point-to-point or point-to-multipoint links. Mesh has become the wrong synonym for metro-scale Wi-Fi/wireless.
Unlike Go and other firms, Wavion says their plan is to sell their chips and ideas to metro-scale firms that will benefit. That means that while their Web site has a Product link that says their AP is in trials, their official top-level goal isn't competition but integration.
Wavion claims a huge reduction in the number of nodes needed while providing higher levels of service. They say they can replace three to four nodes with one of their access points, reducing cost by 50 percent. That implies that their nodes would cost twice as much as Tropos's, as it sounds pretty clear that Tropos is the target here--it's the lowest-hanging fruit with which to compare on the RF front, but it's also the cheapest equipment on the market because of their single-radio design.
Siemens is the vendor of record now for Toronto's massive project, but they'll use BelAir gear: It's almost a footnote in the press release trumpeting Siemens receiving the Toronto Hydro Telecom contract to handle the network building for their city-wide Wi-Fi network. The network will cover 10 square kilometers in the current plan, which will almost certainly grow if successful. Downtown access should be up and running by the end of 2006. BelAir's equipment has been used in other cities and projects, but they haven't yet had a project with this marquee value. When I asked the firm for a good deployment to look at a few months ago, they pointed me to Galt, Calif., a remarkably fast-growing Sacramento/Bay Area exurb that had (at that time) about 80 mesh nodes.
More on the 802.11s mesh standard as word of a January compromise hits trade publications: Apparently, we all missed the news in January and March that progress was made on 802.11s, a mesh standard coming through the IEEE process, because Motorola's announcement a few days ago that they would support the draft and final version of the standard has sparked several articles. At Wi-Fi Planet, Eric Griffith tries to figure out the scope of what the spec encompasses. This can be tricky, because if you're outside the IEEE (or inside and not yet a voting member), you don't have access to the draft itself.
There seems to be a bit of back-and-forth over whether multi-radio outdoor nodes will be accommodated by the final 802.11s. Phil Belanger, who wrote about the indoor, peer-to-peer implications of 802.11s in a white paper published at Wi-Fi Networking News last month, thinks that the impact will be limited to end nodes that can act as mesh devices among each other.
Representatives of Motorola, Strix, and Belanger's former employer BelAir stated in Griffith's article that the implications are broader and that the standard now encompasses outdoor, multi-radio units. But none of the firms would suggest that the standard would, in fact, allow fully interoperable outdoor multi-radio nodes! In the best case, they'd have some minimal level of interoperability that would probably not be terribly useful.
Thus Belanger's original contention seems to be the accurate one, then: in practical terms, it's niche and end-user nodes that will benefit from 802.11s.
Motorola's Mesh Networking Group (Motomesh) has committed to supporting 802.11s: They say their current solution is draft compliant, although the IEEE task group S's notes from the March 2006 meeting states that a merger of two outstanding proposals was successfully made and "confirmed unanimously at the March meeting as the starting point for the 802.11s standard, although much work remains."
Motomesh is confident that their architecture allows over-the-air updates to whatever the draft resolves itself to be. It's unclear why being the only company to be draft compliant before "much work" has been processed on the draft is a selling point over simply stating that the products will be upgradable to the ratified version or later, more fully settled drafts.
As Phil Belanger, formerly of Wayport, Vivato, and BelAir, noted in a white paper published on this site a few weeks ago, 802.11s may have a lot of impact in devices meshing across homes or outdoor environments, rather than any impact on interoperability of mesh nodes that form metro-scale or enterprise-scale networks. It's about the end devices rather than the core mesh nodes, in other words.
Motomesh seems to agree.
Phil Belanger is responsible for Wi-Fi: Not solely, but he's one of the most veteran among industry veterans, and was involved in picking its name. (Note: Wi-Fi really doesn't stand for wireless fidelity or anything at all. Ask Phil.) He co-wrote a document that formed the basis of 802.11, helped found the Wi-Fi Alliance, and served as its first chair. He's worked for many Wi-Fi related firms, including Wayport, Vivato, and BelAir. He recently amicably left BelAir after a decade on Internet time working on startups. As a result, he has been able to catch his breath and reflect a bit.
Phil offers this two-page commentary (downloadable in PDF form) on the future of IEEE 802.11s, the in-progress mesh standards task group. Phil maintains that 802.11s will have little impact on outdoor mesh deployments, and that its real utility will be in providing a basis for home devices to intercommunicate and relay data--every client will be an access point with no extra configuration or fuss, in contrast to Wireless Distribution System (WDS), which is a relatively undocumented and variably supported element found in the original 802.11b spec. Phil notes that more mesh in the house means less signal power is needed, allowing more spectrum reuse and less interference among devices and neighbors.
The 802.11s standard for mesh networking coalesced last week: Two leading groups with separate proposals, instead of slugging it out for a year or two, asked for and were given permission to attempt to merge into a joint proposal in January. At last week's IEEE meeting, the joint proposal was unanimously confirmed as the basis on which to proceed.
What this means is that within a year, there could be a unified standard that mesh devices could conform to for interoperability. I wrote in the Economist this week about one risk to municipal networks' early adoption was that at least four major metro-scale mesh equipment vendors are still categorized as startups. A shutdown or change in direction could leave superannuated equipment scattered like Metricom's or Vivato's.
A standard at least moves towards the potential of a trade group emerging that could set profiles--a la WiMax Forum--for kinds of mesh behavior. There could be single radio, switched multiple radio, contention-free sectorized mesh, and other profiles probably designed by frequency (2.4 GHz, 4.9 GHz, and 5 GHz); it's unlikely there would be a one-size-fits-all. Standards open industries to additional competition, but they can also soothe worried purchasers.
Tropos released version 5.1 of their MetroMesh software, which controls mesh deployments: This release increases capacity, Tropos said in a press release, by using information gathered at mesh end points about the RF environment in which they operate to select or change channels dynamically. In the architecture for city-scale operations that Tropos and EarthLink have described to me, a cluster of several mesh nodes will be paired with each back-haul node. The adjacency of clusters, however, would allow nodes to swap clusters by changing channels or join onto one channel to have more access to backhaul.
The software update also allows edge provisioning, or setting maximum burst data rates for end-users. This is critical for city-wide deployments, where customers will be given no more than 1 Mbps service. The software also ties in authentication with what they describe as "large network operators." Although the text is a bit dense, it describes seamless, session-persistent roaming through secure tunneling, which is akin to Mobile IP as implemented by NetMotion Wireless, for one.