My day began by meeting with a couple of European ARC colleagues to discuss the emergence of “edge control” and ended with me seeing this article on sdxcentral about edge virtualization. Whether your work is in telecommunications, automotive, IoT, medical technology, or industrial automation, today you’re hearing about The Edge. All these fields have (for decades) deployed intelligent systems in their operations. But it seems clear to many folks in these businesses are poised for a major change in systems, the likes of which we haven’t seen (in industrial automation) since the 1970's exploitation of microprocessors to create the PLC and the DCS.
What’s going to make these systems revolutionary is that they will combine the real-time functions of today’s industrial embedded systems (such as a PLC) with the software tech of cloud computing. This sounds impossible and improbable. It’s not. There are several new industrial products on the market that do just this in some way. The question is can it be done in hardware that is both industrial grade and truly commoditized? Product suppliers hate commoditization, but end users love it, because it simplifies substitution and reduces cost. Industrial PCs are probably the best example in today’s market of the value commoditization brings. ARC will be doing more research in this area.
The article in sdxcentral was an interview with Said Ouissal, who is a Founder and CEO of Zededa, a venture stage firm developing virtualization solutions for edge systems, and the original developer of “Project EVE”, an open source edge virtualization project of the Linux Foundation. Ouissal noted 5 areas where edge virtualization differs from that of the data center. These are:
- Zero touch – You can’t reach or even log on to many edge systems
- Heterogeneous – Widely varying types of equipment, software, requirements, etc.
- Cloud native – You can’t economically support a distributed app across dozens or hundreds of remote devices unless it is cloud native – unless the app can be maintained centrally and delivered automatically.
- Security – In a data center there is at least physical security. At the edge there’s often not even that. It’s a zero-trust zone.
- Scale – The edge will feature large numbers of nodes, each running very few apps (perhaps only one). Incremental costs must be very low or growth will be impossible to fund.
IMHO these 5 points are spot on. I would add a double dose of heterogeneity, because of the real-time and reliability requirements of many industrial automation and OT systems. But I’m convinced that the way forward involves virtualizing and containerizing many applications that today run as embedded systems. I don’t see another way to achieve these goals. Do you? Please let me know what you think.