Haystack Technologies: Building a network for the edge

Wander around any tech event and the topic of edge computing is bound to come up. After a decade or so of preaching the wisdom of clouds, the internet of things is forcing companies to embrace the edge.

Dive a little deeper and you realize that many of the edge promoters are selling boxes that sit between a network of sensors and the cloud doing some form of intermediary processing and then sending the data along. But what if the intelligence in the edge really was at the edge? In the sensors themselves?

That’s the pitch that Pat Burns, the CEO of Haystack Technologies, has. He believes that the current model of edge computing can’t scale with the growth of sensors, and the tech world would be better served by data that stays on a sensor until it’s needed.

The idea is that instead of a device always transmitting state to a gateway or the cloud, instead it would wait until it’s queried. Or, if a threshold was reached it would send the necessary data along. But sending the data doesn’t have to rely on direct transmission. The software Haystack uses is set up as a mesh and can run on devices with as little as 30 kilobytes of memory.

His company makes a software layer that sits on top of any radio that can transmit in the sub-gigahertz range. The software layer is part of a standard called DASH7. I’m not a huge fan of proprietary software stacks, but the industry is currently searching for ways to build easy low power wide area networks for the IoT, and Haystack’s tech can offer that.

Because DASH7 works at sub0gigahertz range, the data can travel farther. Burns is currently bullish on running Dash7 over LoRA radios, but it can also work over many others. To me the most interesting element of my chat with Burns was the idea of keeping data at the farthest edge of the network until it’s needed.

I believe we’re heading for such a rapid proliferation of sensors and sensor data that we have to figure out new architectures that let us get the information we need without overburdening our networks. This feels like a tall order. Maybe Burns is working toward the solution.