Yesterday’s FCC wireless technology workshop was filled with some very misguided testimony from Sascha Meinrath, Director of the Open Technology Initiative at the New America Foundation.  Since no one challenged Mr. Meinrath at the hearings yesterday, I felt compelled to set the record straight.  Ethernet and wireless network standards pioneer Richard Bennett who was also listening in on the video stream gave a similar assessment of Meinrath’s testimony saying:

“His comments at the mobile wireless session on European hobbyist networks were quite entertaining, although not particularly serious. Claiming that “hub-and-spoke” networks are less scalable and efficient than wireless meshes is not credible”

What Bennett was debunking was Meinrath’s claim that ad-hoc “peer to peer” (not to be confused with software protocol used to share files on the Internet) wireless mesh networks without any type of centralized and managed access point were somehow superior to traditional “hub-and-spoke command-and-control” wireless architectures that relied on access points, cell towers and network operators.  In Meinrath’s world, users and their unlicensed devices would just magically find available spectrum and connect to other users through other users without any formal hub infrastructure.  Presumably the only thing holding this back is the lack of a spectrum commons where end user devices just manage themselves and broadcast on any frequency at any power level they see fit.  Meinrath also alluded that there must be some kind of conspiracy in the United States because it doesn’t implement any of these infrastructure-less peer to peer mesh network works yet they are fully operational in Europe.

During yesterday’s testimony, Meinrath recited a list of what Richard Bennett aptly named “European hobbyist networks” that had supposedly implemented large-scale mesh networks that spanned metropolitan sized areas.  But Meinrath’s examples are flawed for multiple reasons.

1. The coverage of these hobbyist networks aren’t by any stretch of the imagination scalable, usable, complete, or reliable enough to substitute for commercial wireless services. If they were, everyone would have canceled their wireless network subscriptions and there would have been a revolution in Europe long ago opting for the free open source mesh network instead.
2. Every example raised were in fact hub-and-spoke on the access portion of the network which is the part that actually connects the clients e.g., your laptop to the network via some access point.  These are not examples of the mythical peer to peer client-only wireless network Meinrath is dreaming about.
3. The portions of these “open source mesh” that are labeled “mesh” are almost certainly not ad hoc and they were most definitely set up with carefully aimed directional antennas or else their performance, range, and reliability would almost be useless.  The more appropriate name for this “mesh” portion of the network is fixed line-of-site wireless back haul (or bridge) which is no different than what commercial wireless providers use today where ever appropriate.
4. Many of these ad hoc networks often rely on broadband networks for their back haul and uplink to the Internet (against their broadband terms of service which prohibits the sharing or reselling of bandwidth to the whole world).  This is analogous to a person going to an all-you-can-eat buffet and sharing their plate with 40 other people waiting at the back door.

What is clear is that all of these “examples” Meinrath raised aren’t peer to peer client-only mesh networks at all and they are the very type of hub-and-spoke wireless network that Meinrath claims is so inferior.  The OLPC project for example had disastrous results trying to use client-only peer to peer mesh networks and it unnecessarily drove up the cost and development time while sacrificing battery life.  I pointed out early on that the OLPC mesh project was doomed to failure because a simple $50 router/access point with a $50 antenna would have been infinitely superior because the fundamental wireless engineering and experience tells that mesh is the least efficient technology.  Peer to peer mesh is simply completely impractical for civilian applications anywhere in the US, Europe, or this planet.

The lone exception where peer to peer mesh is practical is military applications where high throughput isn’t required and resilience against enemy munitions is the primary design objective, but this is completely irrelevant to the civilian sector.  But even in this example, the Satellite serves as the primary hub for back haul.  Even if civilians had permission to transmit at high power levels, it is totally impractical for civilian wireless applications unless consumers were willing to carry WWII era radio antennas.  One of the main reasons we use optimally positioned large towers at the hub is to enable the use of small and weak antennas on the client side, yet Meinrath believes we have to get rid of the hub.

From an engineering standpoint, any mesh network is fundamentally going to be many times slower than a point-to-point solution.  We can imagine a cargo fleet of 30 airplanes transporting goods from San Francisco to New York but having to stop in Las Vegas and Denver Airports.  We would have to allocate 10 planes on each leg of the flight rather than all 30 airplanes for one continuous trip from coast to coast.  Even ignoring layover delays and the time it takes to take off and land, we’re still going to lose 2/3 of our optimum 1-hop cargo shipping throughput.  In figure 1 below, I illustrate the fundamental limitation of wireless peer to peer mesh topology.

Figure 1: Fundamental limitations of “peer to peer” wireless mesh networking

In the top example where no repeaters are used, we can use X MHz of spectrum where X is the maximum spectrum available to us.  If X is 20 MHz, and we can achieve 4 Mbps per MHz using 802.11n, then we could get up to 80 Mbps of throughput.  However, the value of X isn’t important here as this fundamental limit of mesh networking applies to any value of X.

In the middle example where we use one repeater R, R can only relay messages when N1 is not transmitting and N1 can only transmit messages when R isn’t relaying.  Even assuming we have perfect frequency sharing, the bandwidth we could use to transmit the file along each hop is only half of what it was without a repeater.  With switching overhead, it’s less than half the theoretical performance.

In the bottom example where two repeaters R1 and R2 are used, the theoretical bandwidth is cut down to a third and actual bandwidth is even worse when overhead is accounted for.  This is because R1 can’t talk while N1 is transmitting or R2 is relaying and that fundamentally divides the available spectrum by 3.

Now we could in theory use multiple radios and perfectly aligned directional antennas so that we only lose half our available spectrum no matter how many repeaters there are, but this really isn’t a “mesh” network because it’s really just a fixed wireless back haul.  Directional antennas are a non starter for non stationary mobile devices because there’s no easy way to maintain alignment.  While there might be some potential for “smart antennas” in mobile devices which use software to determine an optimum antenna configuration that can achieve limited beam forming, their low power levels and tiny antenna arrays can’t compare to large dish antennas with very tight beams.  Furthermore, any mobile devices that had multiple radios to spare would use Multiple In Multiple Out (MIMO) spatial multiplexing to achieve superior throughput before they squander it on a peer to peer mesh network whose weak nodes may or may not even be powered on much less be in decent range.

I spoke with engineer Brett Glass (who was on the second wireless panel at yesterday’s wireless workshop) who founded Lariat, a rural wireless broadband provider about the possibility of using mesh networks.  Brett explained that there is no way he would use mesh because it simply isn’t practical.  The one time he did try mesh involved a single fixed repeater and even that was a nightmare in performance and reliability.  Brett Glass would certainly know a good thing if he saw it and mesh technology is a nonstarter.

To bolster his weak case, Meinrath repeated another common misconception that unlicensed spectrum is only unsuccessful because the power limits for them are far too low compared to licensed radio transmissions.  But this theory is absurd because power limits are the only thing that enables massive spectrum reuse.  With unlicensed spectrum, the number of users on the spectrum is vast which means power limit must be set small enough so that everyone’s signal doesn’t propagate too far.  Even with the current power constraints of 100 milliwatts on indoor consumer grade wireless Access Points, I’m still seeing 6 of my neighbor’s Wi-Fi networks from my home which frequently causes me interference if my neighbors are actively transmitting data on the same channel.  The last thing anyone wants is to have 1000 of their high-powered Wi-Fi neighbors penetrating their home and jamming their own Wi-Fi signal.

Even with existing power level limits, this paper at the UK’s Ofcom found that some metropolitan areas were losing 90% of the available spectrum to beacons and other management frame transmissions leaving only 10% of the spectrum for actual data transmission.  With cell phone tower level limits on unlicensed popular devices like Wi-Fi, we could easily envision a world where 10,000 homes are all interfering with each other to the point where zero useful data can ever be transmitted.  Experience has shown that if anything, power levels for urban and suburban areas are actually set too high.

In conclusion, Sascha Meinrath’s testimony on wireless peer to peer mesh networking is simply not credible because every example given by Meinrath were in fact hub-and-spoke networks.  I sincerely hope that the FCC will review all these facts and disregard these preposterous claims that unlicensed peer to peer mesh wireless networks are superior to licensed hub-and-spoke wireless networks.