The cloud has transformed how we use technology. Of that there is no doubt. But as we have grown attuned to it for the IoT market, complexities arose. Once we moved past the exciting unicorn cases, practical issues reared their head. This has helped bring a renewed interest for the edge, where ever-more adept devices are handling tasks we’d normally assign to the centre, to the cloud.
But this is also an interesting pendulum swing, for we were effectively moving from the edge to the centre and now from the centre back to the edge. Did the cloud fail our overall expectations, leading us to look elsewhere for solutions… or is something else going on here? Will the edge start to replace the cloud? How will the relationship between the two play out?
There are two popular deployment strategies currently in the market, which I can illustrate through the example of machine vision - that being the ability to give a device visual recognition capabilities. One is to have edge devices such as cameras that are loaded with processing power. They can do most of the heavy lifting before the data is sent to the centre. The second strategy is to use dumb cameras that feed data to an intelligent cloud, which then makes sense of the visual data.
Both have pros and cons. Smart edge devices are fantastic, but also expensive and often won’t complement previous investments. The cloud can empower older ‘dumb’ devices, but this comes with latency and massive data costs to consider. It is also not entirely possible to replace the cloud with the edge, not if you wish to tap the levels of Machine Learning (ML) being enabled by the centre.
But I don’t think it’s a case of either. It will be both, and by doing so we can solve a larger problem in the IoT space: interoperability and standards.
Let’s say you have five dumb security cameras. You can connect those to the cloud through a series of steps that layer complexity onto your ecosystem. But we prefer a gateway approach: connect all five to one smart gateway, which then interacts with a single platform. This is what we’ve been doing with the Raptor gateway and IoT.nxt platform.
The platform connects to the cloud, where the development of the ML capabilities live. These can crunch the data coming in, but more importantly, that ML capability can then be assigned to edge devices. Think of it as alpha and beta: the alpha ML algorithm stays in the cloud, but it is exported to the edge as the beta ML algorithm. The edge ML is capable of handling many tasks, deferring that which it can’t to the cloud ML. The cloud algorithm is continually being developed and the edge algorithm is periodically updated to that new standard. then a third dynamic algorithm can eventually make the decision of what happens where. To us observers, it will become a seamless environment.
This relationship is already being bred, evident by ‘AI’ chips being introduced in flagship phones. But it doesn’t even need to be that sophisticated.
