The definition of “wireless mesh network” is a gnarly one.
I have no idea when the term was first minted (best guess: early 1990s), or how many subject specialists scoff when they hear it, but the term is in use, it even has a non-stub wiki, and coincides, overlaps, envelopes, or is confused with several other terms, and understandably misused. Even savvy journalists, during periods when mesh networks enter public discussion, go astray.
We’re amid one such period now.
Interest in wireless mesh networking (WMN) ebbs and flows with public awareness and outcry over Internet reform, because wireless mesh networks—by whatever name—hold potential for changing how the physical Internet is designed, and the public Internet used. They tend to pop up as cursory news items surrounding broader developments or disruptions to network infrastructure and markets, especially in the technical press1.
As the tumultuous FCC hustles to untie the Open Internet Order, Senators and Reps hold hearing after hearing which make abundantly clear that no one, I mean no one, has any slam-dunk idea what the Internet has become, or how to control its increasingly unpredictable behaviors and vulnerabilities.
WMNs are singled out as alternative frameworks to the rigid presence of the major ISPs, sometimes with little regard to what a wireless mesh network might be, or limitations to their use.
What is a Wireless Mesh Network?
Wireless mesh networks (WMNs) are groups of interconnected data radios which send and receive digital information across wireless links using special routing rules meant to improve distributions of traffic across “the mesh,” and/or enable spontaneous, low-cost construction of computer networks more resilient than other wired and wireless designs.
WMNs seem to be a subcategory of wireless/mobile ad-hoc networks known eponymously as W/MANETs (which along with several dozen other long acronyms will be explored and exploded in a future post).
They “seem,” rather than “are” a subcategory because the definitions I dredged up - the one in the prior link from some nerds at the IETF - are not exactly iron-clad. the categories seem to be under construction. It is one goal of this article to contribute questions to the taxonomic certainty of this phrase.
Wireless mesh networks are distinguished from other radio mesh networks like mobile ad-hoc nets/MANETs in that they:
- Are composed of fixed devices, like routers and antennas on building rooftops. These fixed devices are often consumer-grade, often running OpenWRT
- Are usually connected to the public Internet through one or more gateways, vs. being isolated and self-contained
- Often refer to community and research projects of small scales which meet the previous two criteria, like the spectacular efforts of the Red Hook Initiative + Open Technology Institute and NYCmesh here in New York
- Bonus advanced bullet: they alter the IP layer to mesh, as well as RAN: i.e. use ole’ OSI.
Illustration of “unstructured peer-to-peer network” with wireless mesh configuration, courtesy Mesoderm
Wireless mesh networks cooperatively, intelligently distribute traffic across all devices within the network. These technical rules encourage and sometimes demand users of the mesh develop frameworks of ownership and governance which are less centralized, more democratic, and community-based.
In other words, true wireless meshes are antithetical and in some ways, under certain circumstances, superior to the design of today’s Internet, and may suffer less from Pale Horsemen the disembodied Will of the Network of Networks sends our way—even though they’re cut from the same cloth.
A “data radio” can be any, any, type of device with nearly any form of computer (including simple modules and microprocessors) which controls a radio according to a mesh routing protocol, which is connected to a radio as the primary mode of making interconnections and relaying information from one data radio to another.
“Data radio” is not commonly used, but is used here because it covers nearly all devices which are capable of entering a wireless mesh configuration. These devices include things that do not move, that are fixed, like routers, computers with routers inside them or external connections to radios, and special integrated antennas–fixed devices tend to dominate the builds of wireless mesh networks vs. mobile ad hoc—as well as devices that do move, that are mobile, like bluetooth or WiFi or cellular enabled smartphones, laptops, and all manner of embedded devices like cameras, microphones, walkie-talkies—whatever.
- Wireless mesh networks use wireless waves to send information. I offer no apology for how obvious this sounds. It must be driven straight, and true. Wireless mesh networking offers nothing revolutionary if the network is 100% wired. Another way: you can have a network which is a mesh network but not wireless, but you can’t have a legitimate WMN which is not at least partially wireless.
- Wireless mesh networks have superior “survivability” or likelihood they’ll survive adverse events which shut them down, and, at small scales at least, “resilience,” resiliance being a boundary of quality below which service performance is deemed unacceptable by users or applications. Because every machine on the mesh can—if required—perform the role of any other machine a true WMN/MANET is made out of, every machine can be destroyed except for two, and the mesh remains online.
- To further reiterate the revolutionary stuff of WMNs: re-distribution of non-binding roles and permissions across devices can produce more equitable, distributed clusters interlinked by radios. Because they depend less or not at all on coaxial or fiber-optic cables trapped forever in a distance between two points, WMNs theoretically might discover optimal distributions of traffic across radio spectrum with minimal to no human management of the network, beyond the early hand of the Meshmaker.
Ideal mesh configurations are exciting: decentralized, self-sensing, optimizing, and organismal. They interconnect any device located anywhere within range of another wireless device with the same permissions. Visionaries imagine them as self-aware - not sentient - but stupidly self-aware in a manner that might bring breathtaking efficiency to the web, like a colony of ants or bees brings to their Queen, and survival of their species as a whole.
WMNs are really cool in that any community (or small municipality) can build one with off the shelf hardware.
Building a community wireless mesh network using fixed devices that is isolated/self-contained (a kind of intranet) is easy and cheap.
Building a fixed wireless mesh network with a legal gateway to the Internet is more expensive and difficult. The team over at unhosted.org wrote a brilliantly concise outline of what’s involved. NYCmesh also has some great decks and walk-thrus, as well as the Red Hook Initiative.
On small scales, tossing up a mesh is easy stuff compared to starting a larger or wired community or municipal ISP. WMNs lower the cost of infrastructure—fewer cables to run, no digging, construction, and so on. They speed the construction of networks since there’s not much to build, and they spread computational resources across the mesh which gives a warm fuzzy feeling that your router is helping out your neighbors’ routers. Not to mention, they can bypass those ISPs of ill repute or problem networks by routing traffic around end nodes that traverse them, plugging instead into less adulterated and parasitized and surveilled connections to and from the WWW.
Both WMNs and MANETs do not scale: not yet. The largest wireless mesh in the world lives in Spain, with about 35,000 connected.
Centralization, De-centralization, and Distribution
Gradients of centralization and distribution are important concepts in computer networking, and essential to understanding why wireless meshes are plausible alternative models for Internet design.
WMNs are resilient because they are decentralized. #2 depends on #1, and vice versa.
Resiliency (#2) and centralization (#1) are critical concerns to network designers and users alike. If a network is not resilient, it breaks down. If a network is too centralized, there are larger numbers of critical single points of failure. Centralization seems to provoke social unrest as well, consumer/citizens become angry and vocal, and social movements are obscured as riotous gatherings.
An accurate simplification of the gradient of centralization and de-centralization is this diagram on Wikimedia originally from a memo by Paul Baran, published by the RAND Corporation as one of several memos in the legendary series “Distributed Communications.”2
On these graphs, each dot is a “station” or “node,” which represents a computer or server or any kind of device connected to a network. Each line represents one path between two nodes, but not necessarily the physical distance between them.
Each of the three graphs is a network. On the right is a fully distributed network. Information from one node always has more than one path to another node. The network is connected to itself, and can “see” itself by looking up the identities and locations of all nodes on the mesh.
On the left is a fully centralized network. In a centralized topology, all nodes must ask or travel through a single node to access or send information to other nodes. That is to say, all decision-making power in a centralized network topology for deciding how the network lives, and what information is exchanged between nodes, is concentrated within a single node. There is a tenuous consensus centralized architectures are not good, or, not good enough at doing what society wants them to do. However, there are good things about centralized gateways managed by competent humans—the polar opposite of the adminadmin mentality.
Distributed networks as depicted on the right are rare. While centralized architectures as in the leftmost graph don’t exist at all—not for broadband. Those who visualize networks using graphs usually demonstrate the Internet as somewhere in-between the two extremes, and at the present time, skewed toward an undesirable centralization.
The Internet, sometimes called an Internet of internets, is an evolving constellation which distributes and condenses neither randomly, nor with monolithic intent, but to some intermediate point on a spectrum between complete centralization and distribution.
A WMN or ad-hoc network in a dense urban area—which is where the broadband service kinds seem to work best—might have relationships like those depicted in this graph by Rocchini as posted on the Commons.
Because the lines between nodes in this graph also represent distance, you could place a geographic map underneath it for a better understanding.
Some of the nodes might be mobile smartphones, though to reiterate, the infrastructure in a “wireless mesh network,” compared to a MANET, is fixed, as in this photograph of a router running OpenWRT, in range of nearby routers running the same, from the Red Hook Initiative, placed on the Commons courtesy June Butler et al:
While WMNs use IP routing schemes to mesh, the wireless/radio layer is what gilds wireless mesh networks with possibility, and throttles their feasibility at the present time. A meshed node always has multiple options for the path it chooses toward its destination. If a path is or becomes impassable—like when a fiber-optic cable is sliced by a backhoe—a wireless node creates a new path, because it can, using the magic of wireless.
Leading image courtesy Quintanomedia.
This one is bound to whip up a few tantrums from devs, engineers, network designers, theorists, & cetera. They will tell you the Internet IS a mesh network, and that it is decentralized, already. Technically, there are many perspectives where this is true. The biggest scenario is the public Internet, the one we love to hate, is made possible by a mesh network protocol of respectable vintage. Therefore when someone less skilled in the art of networking goes on about how “new and innovative” wireless mesh networks are, that person is at risk of getting NARQ’d on Stack. Which begs: is the question what is a wireless mesh network? spread over a 1,400 word post nothing but an overweight NARQ? I hope not.
WMNs follow similar rules for traffic switching/routing as the mesh networking protocols running on BGP servers. And they use IP. But, old school mesh lookups are looking down the barrel of fiber-optic cables, not across the ether of electromagnetic spectrum. When a mesh goes wireless, it becomes something else. The gritty deets on ASN-BGN meshing is outside the scope of this article, but not the scope of this blog—I will return to the subject, with Max’s help, soon. ↩
Due to the age (50+ years), (de)classification, availability on rand.org, availability on Wikimedia Commons, and that this work was produced for a division of the armed forces (although copyright remained with Rand at time of publication), an assumption is made these images may be reproduced. If this is a false assumption, the images will be removed without question at first request of a Rand representative or rights stakeholder. It is the personal opinion of the author that Baran’s “Distributed Communications” series is an overlooked primary historical source which is tremendously useful to explanations of networking first principles for a general audience, for the public good, and is reproduced with this sentiment at the forefront of my intentions. ↩