Lately I’ve been blogging about the challenges of deployment and support/upgrades for industrial IoT edge systems. It impresses me that one sees wide variation in the support/upgrade model even within the industrial and infrastructure IoT segments (I shall not soil my hands by writing about consumer IoT). When you expand your view beyond just support to the complete edge device/gateway/computer the number of permutations expands exponentially. No problem if you are an academic researcher or (perhaps) a system integrator. But it’s a huge problem if you want to manage a slew of “things” in the field. Here are just some of the major areas for permutation:
Sensor connectivity – how you capture sensor data
Upstream connectivity – how you connect to the cloud (this used to be called backhaul)
Support model – What parts of the edge software do you upgrade and how
Hardware resources (sensing, compute, storage, networking, CPU architecture)
Software stack (OS, support packages, application environments, etc.)
The many required decisions about edge systems are (hopefully) driven by business and application considerations like the expected life of the app, the scale of deployment, the cost of support truck-rolls (a function of scale, location, and support resources). You get the idea. When there is this much variation in this many dimensions it ends up being called “fragmentation” and rightly so. No single cell in this n-dimensional matrix has enough volume to create market momentum.
Earlier this year at the Hannover Fair the Linux Foundation announced the EdgeX Foundry. EdgeX is an open source IoT interoperability framework* that is OS-agnostic (see figure). Dell developed much of this framework themselves as “Project Fuse”, but decided their [hardware] business would be better served if Fuse became a common framework on the industrial IoT edge. EdgeX will be released later in 2017 along with a certification/branding program managed by the Linux Foundation.
EdgeX has the potential to greatly simplify the industrial IoT edge devices by serving as a reference framework. Will EdgeX succeed in this much-needed role? I sure hope so, but at this point I’m frankly uncertain. There are any number of edge solutions coming onto the market that are special-purpose in the sense that they support few variations but they do some things very well. EdgeX can’t be all things to all things (sorry!). But if EdgeX can serve a sizable share of this diverse (fragmented) market well that would represent a major contribution and a lessening of edge fragmentation.
By the way, my process automation colleagues who are now busy trying to develop a new process automation architecture that uses IoT techs would be well advised to put EdgeX on their short list of technologies to evaluate. The OS agnosticism is a nice feature. The Linux Foundation has made noises about EdgeX support for RTOS. And the EdgeX device and core services may be a good model for the Open Process Automation Forum to follow.
It may be stimulating fun to create a one-off “Science Project” in the IoT or in automation, but it is not fun to deploy one or (heaven forbid) to support one.
*I could use the “P-word” to describe EdgeX Foundry. But the “P-word” is so worn out today that it should (like the word “seamless”) be avoided, right?