Good money after bad?

Glowing articles appear regularly in the press about the current state of venture capital funding. As the founder of a small start-up, I have to say that the reality glimmers a bit more feebly. It's not a question of how much money is going into venture firms; it's where the money is being spent.

The San Jose Mercury News recently reported on the ongoing funding story of Visto, a Redwood City start-up. Ten years (and many investors, employees, and board members) ago, the company started as RoamPage. A staggering $267 Million in venture capital infusions later, the company has changed strategies and leaders a few times along the way … and yet it still attracts interest from new investors. Of course, the old investors and early employees have all seen their value washed out and have departed. But new money continues to flow in, including the most recent round of $51 million.

I don't know anything more about Visto, its leaders, or its current business plan than I have read in the paper, but it's maddening that so much venture capital is flowing to companies like this and so little is available for smaller companies like mine with real technology, real customers, and real revenue streams.

I've been through venture rounds with a few companies, so I am not completely naïve. I recognize that the money flows where the VCs can fantasize about a 10X return -- no matter the risks. A normal investor risking their own money would accept more modest returns for an investment in a solid, growing company. But VCs continue to pass on opportunities like my own company for the pot of gold over the rainbow.

As the story of Visto shows, these high-profile, glamorous, sexy start-ups often end up a complete loss for early investors and employees. Yet VCs continue to throw good money after bad. It's a shame that so many of these "big name" start-ups attract only wave after wave of VCs, never the solid paying customers that would create sustainable business and repay the sweat equity of the early employees.

(By the way, we need a lot less than $267 Million, or even $51 Million, to achieve our business plan -- so VCs, feel free to call.)

Francis daCosta
www.meshdynamics.com

Talkin' 'bout my generation

As an engineer, I often assume an idea is self-evident. But clearly that is not true for the concept of third-generation wireless mesh, if the emails I have received about the last post are any indication.

So today, a brief non-technical discussion of the generations of wireless mesh that may make the preceding post more intelligible.

In all of this discussion, I believe it is important to keep in mind two real-world problems. First, in the unlicensed frequency bands (such as 802.11), RF spectrum (and especially use and re-use of available channels) is always a major limiting factor. Interfering radio sources may spring up at any time. Channel and routing agility is the only answer for this unavoidable problem.

In addition, the RF spectrum contention in 802.11-based systems creates throughput problems that can only be effectively dealt with by creating separate contention domains. This in turn, is only possible through the use of multiple logical- or physical radios.

Third-generation wireless mesh solutions incorporate these two elements. And what about the first two generations? Glad you asked.

The first solutions for outdoor wireless networking created access points that could not only connect to users, but also could create links from node-to-node. First developed for military applications, this first-generation product is sometimes called "ad hoc" or "single radio" mesh. This was the first wireless technology to allow some extended distance between wired or fiber connections.

A welcome initial step, users rapidly discovered challenges with first-generation technology. Because a single radio provides both service (connection to individual user devices) and backhaul (links across the mesh to the wired or fiber connection), wireless congestion and contention takes place at every node. Users soon discovered that only one or two radio "hops" were possible between connections to the wired or fiber Ethernet. Support is also very poor for Voice and video applications because of excessive and varying delay across the network. This is precisely what Google is discovering with the first-generation mesh technology deployed in Mountain View.

In an effort to solve the contention and congestion issue, second-generation mesh was developed by placing two radios in each node. This technology separated user traffic from backhaul traffic by creating a separate network for each. This was often achieved by combining an 802.11b/g service radio with an 802.11a backhaul radio in each node.

While this offers significant performance improvement over first-generation mesh, problems remain. With heavy user demand, there is still contention and congestion on the backhaul links, which of necessity still share a single radio channel. This limits the number of radio hops before another costly wired or fiber Ethernet connection is needed. The performance of voice and video applications also suffers because of the congestion in the backhaul stream.

Third-generation solutions add more logical- or physical radios. To overcome the problems of congestion and contention, one radio is used to create a link to its upstream (nearer the wired source or "root") node. Another radio creates a link downstream to the next neighbor node. Unlike the second-generation solution, these two radios make use of different channels. This increases the bandwidth of the network in two ways.



Firstly, each node may be sending and receiving simultaneously to its upstream and downstream neighbors, unlike the backhaul radio of the second-generation wireless mesh, which must continually "turn around" between sending and receiving upstream and downstream. (It's even worse for the first-generation single-radio mesh network used in Mountain View; the node must manage users and backhaul on the same radio!) Secondly, because each link is managed independently, the available channels may be re-used across the network. This expands the available spectrum, increasing performance of the network 50- to 1000 times or more compared to first- and second-generation solutions.

Third-generation wireless mesh solutions also dynamically detect and avoid interference. Each individual node contains the equivalent of a radio spectrum robot, monitoring other radio traffic, tracking its neighbor third-generation mesh nodes, and adjusting the topology and channel mapping of the network instantaneously, automatically, and without disturbing users' sessions.

This third-generation technology is what is needed for major metro deployments, and the problems in Mountain View provide ample proof of that fact.

Francis daCosta
www.meshdynamics.com

It's the architecture, stupid!

So now the early returns are in on Google's wireless mesh network dry run in Mountain View, and they are not good.
http://www.lightreading.com/document.asp?doc_id=109935&WT.svl=news1_3

As the founder and CTO of a company that makes wireless mesh networking equipment, you might think this would distress me. On the contrary, I am delighted. The reason is simple -- wireless mesh municipal networks are finally growing up. And as they do, the expectations change, from "Gee isn't it neat that we can surf in the park" to "Hey, I really expected this to be a replacement for DSL and it sucks!".

The problem isn't wireless mesh, it's the architecture Google chose for the Mountain View network. First-generation wireless mesh networks use one radio for both client service and the backhaul (node-to-node) connections. These networks work great when there is little or no traffic. But they fall apart when activity grows.

Google could have known this would happen. Similar networks failed in similar ways in Chaska, Minnesota and elsewhere. My own company's products are now being deployed in many locations where earlier-generation products, like the one Google chose, were tried and failed. This is exactly analogous to what happened when LAN Hubs were replaced by LAN Switches. It's the architecture, stupid!

As I said, I think this is terrific news. Our industry is discovering that one can't just throw any mesh product out there and have it work. First-generation products fail because of the contention inherent in their architecture. They are doomed to crash when network loads increase. As they ALWAYS WILL.

By contrast, third-generation products use multiple physical- or logical radios, separating client service from backhaul connections. They scale well and can support demanding data, video, and voice applications 50 to 1000 times better than first-generation technologies.

Google has been said to be using Mountain View as a test bed for a future major wireless mesh deployment in San Francisco. So this is a valuable lesson learned. When it comes time for the San Francisco deployment (or to fix what ails Mountain View), hopefully Google will take advantage of the latest architecture to avoid disappointing users, advertisers, and city officials.

Francis daCosta
http://www.meshdynamics.com/