Ad hoc networks, the "next great thing" in the consumer Internet, have hidden roots that run surprisingly deep into military R&D. Although ad hoc networking was born for battlefields, its most far-reaching effect will be in quite a different realm. It is the missing ingredient for the wireless Internet.
Today every advanced military is rolling out large-scale ad hoc networks to carry a variety of tactical communications traffic. The new networks carry all forms of traffic-voice, video, data-using robust, meshed connectivity between wireless routers.
In these networks, many or all nodes may be moving. For instance, the great majority of routers may be mounted in vehicles driving at 60 mph. Some may be in helicopters traveling at much higher speeds across a battlefield. The networks are thus very dynamic; links between routers appear and disappear in a matter of seconds. Such ad hoc networks are up and running in the United States, Canada and England under their respective militaries.
The basic technology needed for ad hoc networks is best explained by network layers. At the physical layer, some form of packet-based wireless transceivers is required. In principle, these may be based on a wide variety of physical channels-RF or optical, or even acoustic or magnetic in specialized circumstances. In practice, RF is generally employed. In commercial settings, the IEEE 802.11 radio family is a nearly universal favorite, running at 2.4 or 5 GHz, although some companies employ a proprietary radio. Bluetooth and other low-power radios are also used. Military systems generally call for advanced radios with improved transmission security, lower probabilities of intercept and detection, and anti-jam capabilities.
Specialized ad hoc network protocols run above the physical layer and implement a range of mechanisms peculiar to ad hoc networking. Neighbor discovery allows a router to determine which other radio routers are currently within range. Broadcast "beacon" transmissions are usually employed to aid in neighbor discovery. Topology control determines which of all potential links to neighbors should actually be used. In systems with directional antennas, this mechanism may include beam-steering decisions. Link characterization provides a means for determining when a radio link is of sufficient quality to use for data traffic, how good it is compared with other links (for example, how reliably it can deliver packets) and when it should be removed from service. Ad hoc routing determines where distant nodes are in the network, to some degree of accuracy, and the hop-by-hop paths by which datagrams may be sent to those distant nodes. Forwarding mechanisms actually relay packets onward through the ad hoc network.
Finally, a variety of Internet-level services must also be provided. These include admission control, provision of Internet Protocol addresses (for example, by DHCP or other means), mapping between IP addresses and link-layer addresses, name servers, exit routers to other portions of the Internet, network-management systems, public-key services for secured networks and so forth.
Spurred by today's military successes in ad hoc networking, and by this technology's obvious potential outside the military, dozens of academic research groups across the world have recently entered the field. A hand-ful of commercial companies have also joined in, often with technology originally developed for the Defense Advanced Research Projects Agency (Darpa).
Academic research has mainly focused on the routing aspect of ad hoc networks, with two competing lines of approach-proactive routing and reactive routing. Proactive techniques attempt to determine the location of nodes in the network at all times, although perhaps not with perfect fidelity, so that traffic can be readily forwarded to these nodes when the need arises. Reactive routing, by contrast, attempts to find a node's location only when traffic is being sent to that node, usually employing some form of flood search to find the location. A further distinction is that proactive routing schemes continually reevaluate the paths by which traffic flows, to adjust them to the changing circumstances, whereas reactive schemes generally stick with an existing path until it can no longer be used, even if it becomes far less than optimal as the network nodes move.
Ad hoc networking's most important role is likely to be as the enabler for the wireless Internet. RF spectrum is finite, but each wireless Internet user needs as much of it as possible for high-speed connectivity. This requirement leads inexorably toward a very large number of basestations to minimize the number of users contending for a shared basestation antenna. In turn, this leads to ad hoc networking, whose self-organizing, self-healing properties minimize the planning and maintenance needed for large radio networks.
Two demands in the wireless Internet space are very clear: standards and scalability. Both are ultimately network protocol issues. The need for worldwide, open standards has been the path to all networking success to date. But, remarkably, it hasn't yet been tried in the wireless Internet. Here the Internet standards are essential, and this means the standard set of Internet protocols, not a version tailored for wireless users. The lower layers of the protocol stack will almost certainly be IEEE 802.11 in one form or another.
A variety of new techniques is just emerging from the lab, all distinctly useful for the commercial wireless Internet as well as the military. Within the past year, Cesar Santivanez and Ram Ramanathan, both of BBN, have provided the first theoretical understanding of exactly how ad hoc routing protocols scale. This understanding led to the development of "Hazy Sighted" routing-a novel, easy-to-implement link state variant that proves that hierarchy is not mandatory for scalability. In fact, it appears that Hazy Sighted routing scales far better than previously known techniques. This is a fundamental result with immense practical importance.
See related chart