More reasons IoT depends on the edge

Last week, I wrote about the five reasons edge computing is so essential for the internet of things. I focused on security; IP protection; resiliency and latency; bandwidth costs; and the need for local control (if we want to make machines autonomous). But several readers told me that I missed some important considerations.

They felt, for example, that I didn’t spend enough time discussing costs for bandwidth and storage. Another reader said that I missed out on the potential for power savings that accrues when a connected device maker utilizes a user’s device for computing.

Christopher Dow, chief technology officer with August Home, the connected lock maker recently acquired by Assa Abloy, explained that while computing costs are continually going down, the cost to power computers isn’t. With that in mind, it makes sense for any business providing a connected device or service to try to offload as much of those costs on the user as it can.

In August’s case, this means handling computing tasks associated with media playback on a user’s phone as opposed to organizing the video stream from the cloud. It also means that video from the August Doorbell is stored in the device RAM, and is pulled from that RAM if and when it’s needed. Doing it this way saves on storage costs, because it means August is taking advantage of the user’s computing power and electricity rather than paying Amazon Web Services for it.

As Dow illustrates, and others have also explained, costs associated with storage and processing can add up. So while I spent a whole section on the bandwidth costs, I didn’t talk about what it meant to keep all of that data sent up to the cloud, tucked away in an Amazon S3 instance. In many cases, customers who use edge analytics dump a lot of that data once they’ve performed the major calculations on it.

For example, Patrick Lazar, the VP of engineering at FreeWave, a company that sells radios and computing platforms for edge networks, says that one of his firm’s customers cut 80% of their cloud costs by sending only the changed data to the cloud instead of all of it. So if a sensor measuring temperature stayed the same all the time, that data was deleted and only the abnormalities were uploaded.

And finally, there’s a privacy angle, too. While related to both the idea of security and of protecting IP, privacy is worth a separate mention because it matters a lot in the consumer world, especially in the post-GDPR world that will arrive after May 25th. For example, in the wake of the Cambridge Analytica scandal, reports say that Facebook will delay offering a connected consumer device. And when it does offer such a product, it will process certain things like camera images locally so as to soothe privacy worries.

Worries about data collection and sharing already leave many smart home users unwilling to buy cloud-connected cameras or smart speakers, while misuse of data or poor security practices only prove that consumers are right to be concerned. Leaving as much data on the device as possible could help sell products in the future. Apple’s certainly banking on that.

So perhaps I should have titled last week’s story “8 reasons the edge is essential in the internet of things.” Or maybe someone will offer me two more reasons and we can make it an even 10. Regardless, companies need to think carefully about why they want to send data to the cloud and whether or not it actually belongs at the edge.