[atinchandel]

With my experience of working on +300 IoT projects - In majority IoT application edge computing is not required, devices generate very less data and perform simple computing or no computimg at all.In embedded usually it's better to keep things simple so that cost and bugs can be reduced. Most IoT application or solutions are usually not adopted due to hardware cost (capex) so keeping devices processing, memory and development cost low is important. Few measures can keep IoT deployment secure and you don't need a complex framework - TPM, Secure Boot, Secure Update and SSL/TLS are enough. I still see that such IoT framework will only be used in small % of IoT applications. With public networks getting more reliable and cost effective edge computing would become more of a maintenance nightmare than benefit.So keeping IoT devices thin and doing most of the processing on cloud is better. Though I do agree that in some application IoT endge analytics and computing has value. I believe we are living in a work of political and technology hype :).

[anonymousDan]

What if you have poor/intermittent backend connectivity, low latency requirements, or are handling sensitive data and don't want to ship it to a cloud backend? I think there can be a place for this model. Perhaps something like GDPR can make it more attractive? In the projects you have worked on, was the primary motivation maintainability or a business need to gather all data centrally in order to perform some kind of global cross device analysis/aggregation?

[atinchandel]

Dan I agree there is place for this model especially when you want to make local decisions or produce lot of data (e.g. video). In most of the projects we handled data required central aggregation.

[veyalla]

Using IoT Edge as a gateway is a popular and supported scenario: https://docs.microsoft.com/en-us/azure/iot-edge/iot-edge-as-...

[nielsole]

If I would design a service I would rather trust my cloud provider with an SSL connection in between, than a heterogeneous mix of servers that may be physically compromised (there are very few places that could house an "edge" at datacenter security standards).

[pletnes]

I work on an image/video analysis type project. Internet connectivity is over (sometimes flaky) LTE. We could not live with a «thin device» unless we had a fiber link to the cloud.

Even if you do simple point measurements (e.g a weather station with temperature, pressure etc) it might make QA sense to perform multiple measurements and reject outliers, take an average etc before reporting to The Cloud.

[anonymousDan]

Is your project/business public? If so any chance you could provide a link? I'm interested in this space from a research angle and it would be great to be able to point to a real-world use case.