RNRT AIRNET 2006–2009

AIRNET website

The aim of this project was to study how to design and build an interconnection infrastructure resting entirely on wireless local networks operating in a public frequency band. The motivation came from an observation: 802.11 (Wi-Fi) networks were being deployed steadily in densely populated areas, offering access to the global Internet through access points. That deployment was spontaneous, and for the time being there was neither coordination nor collaboration between covered areas — access points belonging to different operators, for instance, did not talk to each other. Imagine instead an interconnection infrastructure in which a new access point dynamically joins the set of existing ones and takes part in multi-hop packet relaying to extend coverage. The building blocks of such an architecture can be Wi-Fi access points in the traditional sense, but also mobile hosts willing to play an active part in communication, as in ad hoc networks. Much as Ethernet contributed to the spontaneous growth of the early Internet, we believed that dynamically interconnecting wireless access points could change the communication paradigm profoundly: the spontaneous wireless infrastructure becomes the core of the interconnection network and offers users an ordinary mobile communication service.

Project website.

The project faced several challenges. The main one lies in the characteristics of the radio medium, which behaves very differently from the wired links the existing Internet was built on. Even though wireless networks perform reasonably well compared with mobile telecommunication networks (GPRS, EDGE, UMTS), that performance degrades quickly in a multi-hop infrastructure running standard Internet protocols (TCP/IP). For these reasons we argued that an all-wireless interconnection has to rest on different principles from today’s Internet, whose protocol architecture was not designed to account for wireless characteristics or host mobility:

• The wireless medium imposes new constraints on the communication architecture: signal attenuation, multipath propagation and time distortion cause large variations in throughput, error rate and delay. Spatial position relative to an access point also raises hidden and exposed host problems. The Internet was not designed to handle these constraints efficiently.
• The routing protocols, naming and administration principles of today’s Internet assume a two-level hierarchical core deployed and administered by highly qualified staff. To design a spontaneous wireless interconnection in which potentially every user can take part, a radically different approach is needed, founded on self-organization and self-management, allowing incremental growth without loss of overall capacity.
• The addressing scheme and routing structure of the wired Internet are rigid: they were designed for a fixed infrastructure connecting stationary hosts at a single point of attachment. The solutions proposed for mobility (Mobile IP, in both IPv4 and IPv6, and refinements such as Hierarchical Mobile IP) are stopgaps that add functionality without really solving the problem. A general solution for mobile hosts calls for completely different addressing, naming and routing structures.
• TCP, the most widely used protocol on today’s Internet, does not perform satisfactorily in a wireless environment. A new interconnection architecture has to take that into account.
• Given the variety of wireless technologies, and to allow migration paths from existing radio layers, a unifying routing layer is needed. Dynamic multi-hop ad hoc routing protocols are one candidate technology for such an infrastructure, though extensions are required within that unifying framework, particularly for scalability and quality of service.

In our vision, public wireless access points and mobile hosts cooperate to maintain a wireless communication infrastructure with greater transmission capacity and flexible support for host mobility. The project set out to study how to reach that vision incrementally, as the wired Internet did in its early days, with a minimal set of rules and protocols.