The company that will provision and run the 1,500 square-mile wireless mesh network in Silicon Valley that was announced rrecently is an almost unknown, privately held network operator based in Grand Haven, Michigan, home of the world’s largest musical fountain.
Azulstar Networks has been operating in the US for just three years. Its founder and CEO is a 34-year-old former Intel engineer, Tyler van Houwelingen, who credits a former girlfriend in Spain with coining the company’s name (‘azul’ is Spanish for blue). Grand Haven is van Houwelingen’s hometown and it’s where Azulstar created what it says is the first wi-fi mesh net in the United States. In the Silicon Valley deal, Azulstar will operate the network, IBM will be the main integrator and Cisco will provide the equipment.
Also in the partnership is SeaKay, a local non-profit advocacy group. The Silicon Valley network, covering 42 communities, will offer voice, data, and video services to about 2.4 million residents, which is 28 times the combined populations of Grand Haven and Rio Rancho, New Mexico, where Azulstar also runs a wi-fi mesh. Van Houwelingen says he anticipates no problems. He spoke to John Cox about the giant deal.
Do you have any experience with large-scale mesh nets?
[Yes,] In Grand Haven and Rio Rancho. There are thousands of subscribers and tens of thousands of users of these nets. They’ve been our alpha and beta tests for our [future] very large regional network deployments, which will use equipment that wasn’t available a few years ago.
How big are they?
Grand Haven was the first blanketed wi-fi city, initially covering seven to eight square miles. It’s been extended to surrounding areas and now covers about 25 square miles. Rio Rancho, with about 70,000 population, is about 50 square miles of coverage.
But the Silicon Valley network is vastly bigger. Will the Cisco equipment scale to that?
The cities on the peninsula are next to each other. And each one [will have] their own bandwidth. We’ve been testing the Cisco [1500 outdoor mesh access point] for a year.
We’d been wondering where Cisco was in this whole [municipal wi-fi] game. Then they arrived with their 1500 platform. There are some [mesh] alternatives that I can take down with just my laptop. Cisco stops a lot of that right at the AP [access point]. It’s really a carrier-grade system. And there’s a lot of bandwidth available. You just keep overlaying networks on multiple frequencies. We’re supporting multiple [5 GHz] 802.11a nets, as well as the 4.9-GHz licensed band and in the future, the 5.2-GHz licensed band.
We’ll deliver as much bandwidth using as many of these frequencies as possible. I’m not anticipating any problems with the Valley project.
This won’t be a single network? Each community will have its own?
In our existing nets, we use 5 GHz 11a and pre-WiMAX radios. Users access the net on 2.4 GHz, then the traffic jumps up to the 5-GHz band, nearly doubling your capacity. And recently, another 200 MHz of spectrum was added to this band. If you keep overlaying bands, you can bring more and more capacity to each node. Roaming between nodes is all done on the wi-fi net, with backhaul on the 11a band. We can just keep overlaying these connections.
How can you take down a mesh net?
One way is to take two client wi-fi bridges, for example from Linksys, and then connect them to each other with a crossover cable. That creates a loop in the Ethernet, one of the worst problems for an Ethernet network.
But some vendors, like Tropos and Cisco, can handle that kind of stuff. In our tests, we found one way to take down the Cisco mesh, and they have since fixed that.
How are you going to manage a network that big? Cisco is offering its Wireless Control System, but will that be able to handle tens of thousands of mesh nodes?
We’ve seen what WCS can do. It can do a lot of stuff — like auto-detecting interference — on its own, while optimising bandwidth. It’s still early in the lifecycle of the Cisco gear. But what we’ve seen in the 1500 is a big step forward.
Who pays for this network?
IBM Global Finance is the key to financing this. They’ve come out and said they estimate it will cost [between] US$75 million [and] US$240 million [NZ$376 million]. But US$240 million is way over what we think it will cost.
We’re working with municipalities to [convince them] to use this net for [government applications like] video surveillance, traffic coordination, voice, and so on. All of these are being done elsewhere [in other municipal mesh nets]. The goal is sharing the risk with the public sector. If we get enough government uses on this, the risk [and cost] is shared and lowered. IPTV is one deliverable for this as well. There are a lot of ways to monetise this bandwidth.
Forty-two cities came together for this. The power of the network goes up with its size and the number of users. For example, police cars can connect to this network uniformly across the peninsula, not just in one or two communities.
I can’t imagine negotiating with 42 communities. Have you started talking with them yet?
We’re starting that next week. We’ve been elected to be the preferred vendor to negotiate with the governments.
Different cities will want different things. So we can have pretty accurate data to support location-based advertising. This will be a significant revenue source.
So you’re anticipating several streams of funding, tax dollars and subscriber revenue being two. What else?
Advertising support. We’re talking to potential advertisers. The network is location-based and to access it, users have to register a username and password. So we can pretty [much have] accurate data to support location-based advertising.
Did the request for proposals mandate specific service plans?
The RFP recommended different service level agreements: best effort; premium; indoor service, and so on.
Our pricing was very close to that, and we dramatically bumped up the speeds of the free service to 1Mbit/s download [and] 60kbit/s upload. That’s a fairly slow upload. And you can’t run VoIP or VPNs on the free service.
To get those kinds of services you have to upgrade to one of the commercial services.
What kind of price range will those services have?
At the low end, about 1Mbit/s with all your ports opened for voice and other services, and you get jumped to a higher SLA with separate VLAN and higher priority traffic built in to the net end-to-end, it will be about US$12.95. From there, you can jump to 2Mbit/s and 3Mbit/s services, and T-1 replacement services. These are [estimated] retail prices. We are the wholesaler, selling packages to companies like NetZero and others. We’re having lots of discussions with them.
When will construction start?
I think it will take a few months. We’ve got to get co-location agreements with the various local electrical utilities [and] negotiate rates for unmetered electricity. And then there are the specific negotiations with each municipality over their specific requirements and services. Hopefully, in three months, we’ll be moving stuff.
Isn’t the Valley already well served by existing broadband net providers?
Our price-performance will be very attractive. For low-end users, below 3Mbit/s, we’ll be pretty competitive. And you get the [wi-fi] roaming across the Valley; with 42 cities people are covered all day. And this is bundled [in] with the basic service.
What we see, generally, [is that] you [first] get hundreds of users on your free service, then [see] a trickle of them upgrading to the fee services.
We want people to think of this as their local wi-fi utility.
You do a lot of travelling. Are you using wi-fi networking?
I travel quite a bit. I use whatever wi-fi connectivity is available through iPass [a global mobility provider for securely connecting mobile enterprise users with their corporate net]. It’s not one of our services, but I’m a registered iPass user.