Nubix is shrinking containers to drive growth of what is expected to be a multi-billion-dollar IoT and edge market.

The company released a Developer Edition of its Nubix.io container platform that is targeted at supporting smaller container deployments and smaller services at edge locations. Nubix CEO Rachel Taylor in a blog post explained that the company’s focus is on the extreme edge of the network. “At Nubix, we define edge in an absolute sense, wherever the data is being generated by the sensor/device is the edge, no transport required,” she wrote.

Taylor explained that Nubix’s containers are 100-times smaller than the size of traditional Linux containers like Docker, “which means we can embed our containers on low-power, low-footprint devices.” She noted the container and platform can be as small as 32 kilobytes.

The Nubix containers are self-built “in order to scale down to the smallest of microcontrollers,” explained Tom Yates, VP of products at Nubix, in an email to SDxCentral. They run on Linux, but are architected to be independent of Linux and Linux container technology, he noted, adding that this allows them to run on ARM M-class processors that are typically memory constrained and lack a memory management unit (MMU).

The Developer Edition launch includes an application development platform for those container and application deployments. This provides pre-packaged, drag-and-drop components to build and test applications.

It also uses pre-packaged components and containers to support over-the-air, container-based deployments with hardware abstraction. This includes deployments on Raspberry Pi devices (3B and later) and the BeagleBone Black hardware platforms.

This support plays into what Yates explained was the company's specific focus on the “device edge – the tiny microcontroller-based devices where data is being generate, be that motor controllers or lighting systems or things we haven’t even thought of yet.”

Code written to those devices is independent of the operating system, which reduces deployment risks and speeds updates. “Our containers also provide a hardware abstraction layer that allows for dynamic updates without the potential of bricking the device,” Taylor added, which should alleviate a common fear when dealing with hard-to-reach edge devices.

Nubix is also providing IoT devices with native analytics capabilities that reduce latency, bandwidth, and connectivity issues compared with having to stream that information to a centralized or cloud-based platform for analysis.

"The device edge needs to curate the data sent to the next level where a next level of analytics can be done comparing systems as a whole,"  Yates noted. "To that end, Nubix is investing in developing tiny services that IoT developers can use to help their devices analyze, collate, and curate data at different levels in support of the next-gen enterprise architecture."

Yates also explained that the company was “monitoring” the container landscape, with current evaluations of how the Open Container Initiative (OCI) might support greater interoperability with other container technologies and orchestration systems like Kubernetes.

The Nubix.io platform joins a number of similar projects in the market that are trying to shrink down container deployments to fit into resource-constrained environments at the network edge.

One of those is Balena.io, which uses Linux-based containers and cloud technologies to support IoT deployments. Those efforts revolve around separating core operations of the IoT device from the application layer. This allows for a device to continue to communicate with its network host should an application fail.

The company, which last year changed its name from Resin.io, can manage multiple containers with one running a supervisor agent to make sure the device is running properly and can connect to its platform. The other containers run the user’s applications.

Another similarity is with Zededa, which offers a software platform that can be deployed on virtually any edge device in support of IoT. This involves a lightweight software stack that can adapt to different deployment models.

Yates said that those offerings “share some common market messaging” with Nubix, but that his company is different in its explicit focus on those small edge devices generating data.

The Nubix.io analytics capabilities are similar to what SWIM is doing with its EDX edge software platform. This platform ingests real-time edge data to construct an edge fabric. That edge fabric then applies the company’s analysis and machine learning capabilities to produce a “digital twin.” This twin is used to learn from the data coming into the system to find hidden patterns and predict future behavior.

However, the SWIM platform needs at least two megabytes of storage space, which is significantly more than what Nubix is claiming for its platform.

UPDATE: This story has been updated with more comments from Nubix on how its platform works and how it's differentiated in the market. It's also been updated to note Resin.io's name change to Balena.io.