July 12, 2004, 10:21 AM — OK, this column is going to be a bit esoteric this week -- at least at this point in time. But I've recently published a white paper that's a product of a number of conversations I've had recently on how to architect wireless networks -- particularly those based on mesh architectures, and particularly those that fall into the category called micromeshes.
If you'll recall my last column on the subject, mesh-structured networks have a number of very interesting properties, most notably ease of installation due to the reduced requirement for wired connections to every node (called backhaul). Meshes can relay over the air using other mesh nodes to forward information and provide connectivity to the rest of the network. A micromesh is a special class of wireless mesh network usually designed for sensing, telemetry, control, monitoring, and related activities, and nodes in these networks are typified by their small size, low cost, long battery life, and reliability. They usually don't have much range (transmit power is throttled way back to contribute to long battery life, which may, by the way, be measured in months or even years). Similarly, the amount of processing performed by each node is also quite low for this same reason, and thus we have a bit of a dilemma. Each node in a micromesh must interface to a sensor, actuator, or other similar element. It must act as a relay point for other micromesh nodes. But it's not going to have much storage, processing power, or local software.
So, how do we implement all of the other management, control, database, and application functionality in large micromesh networks? What about security? What about providing a common API that allows mesh nodes to be mixed or upgraded transparently when appropriate and desirable? These questions are part of a larger range of issues that together represent a fundamental challenge to the deployment of micromesh-based apps. As you might guess, micromesh systems are fairly rare these days, but I think they're going to be quite common, even in residential applications, in the not-too-distant future.
Fortunately, there's a time-honored mechanism for addressing the issues raised above: the gateway. Gateways are traditionally a transport-layer (Layer 4) mechanism for moving information through heterogeneous collections of networks. They can also correct for many of the inadequacies in micromesh (and other network) systems. There isn't enough to give you the whole story here -- for more information, see the white paper noted above. And keep in mind that this architecture generalizes quite nicely to other mesh-structured solutions -- and beyond. So, maybe this topic is not so esoteric after all.