Canonical has published an article by Gabriel Aguiar Noury showing how Ubuntu Core 26 works alongside the two big cloud IoT platforms: AWS IoT Greengrass and Azure IoT Edge. The premise is plain enough. Not everything can happen in the cloud, so it makes sense to split the work between the data centre and the devices out in the field.
A hybrid cloud-edge architecture
The approach pairs both sides. The cloud handles central management, model training and fleet coordination. The device captures data, runs inference locally, and only sends the result back. That cuts traffic, lowers latency, and keeps sensitive data from having to leave the hardware.
Ubuntu Core sits underneath all of this. Both the AWS IoT Greengrass runtime and the Azure IoT Edge runtime are delivered as snaps that install on top of the system. That agent snap authorises, authenticates and integrates the device into the end-to-end workflow, so you don’t have to touch each unit by hand.
Deploying AI on the device
Canonical’s example is concrete: a device that captures images and runs real-time object detection. The unit takes the picture, runs the model, and ships only the result back.
What changes is how the logic gets packaged per platform:
- On AWS, the work (image capture, model processing and cloud communication) is split into components.
- On Azure, the same task runs as containerized modules delivered through IoT Hub, with the option of using Intel OpenVINO for the inference part.
Zero-touch onboarding and safe updates
One of the points the post stresses is zero-touch onboarding. Devices can ship from the factory with the snaps already installed. On first boot they connect to the network, authenticate against the cloud platform, and register themselves. No technician needs to walk unit by unit.
Maintenance follows the same pattern. Changes deploy centrally and push out to devices. Here one of Ubuntu Core’s most useful traits comes into play: updates are transactional. If something goes wrong mid-update, the system can roll back to a known good state instead of leaving the device bricked. That matters a lot when you run hardware spread across sites with no easy physical access.
For day-to-day operation, you can read logs, monitor communication with the cloud, and redeploy components remotely. The whole fleet operation runs without setting foot where the devices are.
Why it matters to you
If you work with embedded or IoT devices, this approach answers two familiar headaches: how to update thousands of units without breaking any, and how to run AI close to the data without depending on a perfect cloud connection. Delivering the AWS or Azure agent as a snap fits Ubuntu Core’s model, where the whole system is confined and updated through atomic packages. Coming from the Ubuntu page and interested in the local AI side, you can also look at how Ubuntu Core 26 runs as a local AI appliance in a VM.
Source
Original article by Gabriel Aguiar Noury on the Canonical blog: A look into Ubuntu Core 26: Cloud-powered edge computing with AWS IoT Greengrass and Azure IoT Edge. Ubuntu and Ubuntu Core are products of Canonical.