Scaling Up: An overview of large mesh networks
Wireless mesh networks are being used in a variety of ways around the world and each implementation offers an opportunity for us to learn. There are networks large and small that meet a variety of needs, from sharing Internet access to transmitting sensor data to serving as a neighborhood intranet. With the recent focus on resilient networking technology, a common question we get is, “How well do wireless mesh networks scale?” To answer this question, we’d like to offer an overview of community wireless mesh networks that have been particularly successful, with a focus on those that have scaled to large sizes.
Freifunk operates in several German cities and provides tools for organizers to set up mesh networks in their own towns. Organizers sometimes collaborate with local governments to build permanent mesh networks that offer public Internet access. Individual networks maintain maps and databases, like this map of the Leipzig nodes.
Guifi.net is an online community promoting and tracking mesh networks in and around Spain. It has around 19,000 nodes at the moment, and its own peering point with fiber backhauls. Nodes run the dd-guifi firmware and use BIRD and quagga for routing. Guifi also maintains a map of nodes and routes.
Athens Wireless Metropolitan Network (AWMN) was started in 2002 and provides Internet connectivity throughout cities in Greece. It includes over 1100 backbone nodes and over 2400 client computers. It has expanded widely during its lifetime and is managed in part by a non-profit association. It uses BGP and OLSR for routing.
WasabiNet is a neighborhood mesh network in St. Louis, Missouri. It provides limited free internet access and paid access for different speeds. It uses ROBIN firmware and OpenMesh for short distance links and various platforms for backhauls. See Wasabi’s cloudtrax map for a snapshot of its network topology.
Each of these networks offers valuable connectivity to particular communities while expanding open and neutral Internet access. They have evolved based on community needs (rather than potential profit) and are often quite resilient and flexible. Many of the networks we mentioned above have grown into locations that previously lacked Internet access.
Strengths and Challenges
These networks didn’t set themselves up: network administrators architected almost all of the major backhauls and did extensive planning. The links require regular maintenance by volunteers, and uplinks to the Internet need to be purchased and periodically expanded as networks grow. Mesh routing protocols help with the reliability of a community wireless network in the same way they help with the Internet’s fiber backbones, but at a certain point, engineers are needed to solve problems.
Given the requirements for hardware replacement and technical maintenance, community wireless mesh networks need resources. In some instances a network might receive large grants, while others receive mutual aid from hobbyists and other networks or are bolstered by research projects. By the time a new community wireless mesh network reaches a large scale, the community must take on more responsibility by making significant investments in maintaining and improving it. No large mesh network could survive without financial investments in both infrastructure and people.
In addition to hardware and radio engineering issues, community wireless mesh networks often run up against problems with the routing protocols themselves. Single-radio mesh is far from ideal, but it’s often the only option for small networks due to economics. And larger pools of mesh participants result in all sorts of funny network and information theory questions to answer. Thankfully, there is an active community of developers tackling these issues, and multi-radio/multi-frequency hardware is becoming more affordable.
But before you get too bogged down in reasons why our networks have a hard time scaling, take a moment to reflect on the reasons why we run these networks: are we attempting to simply replace the Internet as it currently stands or are we working toward new, alternative systems?
For further reading, see these resources:
- - - -
This post is part of an ongoing series about our work on the Commotion open source software project. The project aims to use mobile phones, computers, and other wireless devices to create decentralized mesh networks.