From the Magical Mystery Tour to Project Ocre: My Open Source Journey
By Jason Shepherd, CEO of Atym
The year is 2015. Taylor Swift was touring in support of her 1989 album and I didn’t know a lick about open source software. Fast forward to today and I’ve had the pleasure of working with the open source community on many projects. I’m a big believer in the importance of open, trusted infrastructure and am excited to announce the upcoming launch of Project Ocre in the LF Edge community. Read on for more about my journey with the open source community and for details on what we believe is the next frontier of edge computing.
Having helped bootstrap Dell’s emerging IoT business in 2015, I was leading software strategy and partnerships to augment our new line of industrial-grade edge gateways. I was spending a lot of time in the Valley doing what I called the “Magical Mystery Tour” – i.e., driving around to meet with various different companies with IoT platforms. This was at a time when most IoT software had been given a facelift from the M2M days, often monolithic in nature and long in the tooth. I quickly realized how fragmented the market was, and that the last thing the world needed was another horizontal IoT platform. The running joke then was that there were 150 IoT platforms (and then it was 300… then 400…)
Working with the Dell CTO team, we realized that the right architecture for IoT solutions at the edge had mechanisms to bring together diverse application components, update them on different lifecycles, and reduce the tight coupling to underlying hardware. Basically, much like software was being built in the cloud. In late 2015, I helped spin up and co-lead an effort called “FUSE” (named due to “fusing” different solution elements together) within the Dell CTO organization. The intent was to bring cloud-native principles (e.g. microservices, platform independence, continuous delivery, etc.) to IoT edge infrastructure. With our initial prototype completed in early 2016, we started looking at options to bring it to fruition in the real world.
We considered whether Dell should launch an IoT platform, but quickly decided that it didn’t make sense for Dell to be yet another in the sea of offerings. So, we decided it made the most sense to open source the framework to try to drive a standard. After all, standards drive scale.
Our next step was to find a vendor-neutral organization to host the project and facilitate adoption. We evaluated several organizations over the following months and ultimately decided on the Linux Foundation (LF). At that time, myself and many others incorrectly assumed that the Linux Foundation was only about Linux. In fact, the Linux Foundation was the host for the budding Cloud Native Computing Foundation (CNCF) among many other projects.
With the help of the LF, Dell ran project incubation meetings for several months while recruiting other backing companies. We launched EdgeX Foundry in April of 2017 with fifty supporting member organizations. Thanks to the great work of many individuals and companies, EdgeX Foundry is now a globally-used framework for IoT solutions.
While chairing the EdgeX governing board in late 2018, I helped the Linux Foundation bootstrap LF Edge, in collaboration with the EdgeX community and other founding projects such as Akraino, Fledge, and EVE. The idea was to bring together a collection of projects that were building plumbing and facilitating interoperability across the edge continuum.
Meanwhile, in my new role as CTO of IoT and Edge at Dell Technologies, my team had turned our focus to the software infrastructure layer. The fundamental question was the necessary considerations to shift cloud-native workloads out of secure data centers to edge nodes in the wild. This could be the factory floor, an equipment closet in a retail chain store, or on a truck in the middle of nowhere. Many of my colleagues that were focused on traditional data center virtualization believed they had these needs covered, but my team felt that it required more of a grounds-up approach.
Compared to servers in a traditional data center, you have to assume that nodes in distributed edge locations will periodically lose connection to centralized resources, can have no defined physical or network security perimeter, and that IT skills will typically be hard to come by. Pricing models also need to be different as hardware becomes more resource-constrained and less expensive, but much higher in deployed volume. Over the course of several months, our team built out a vision for extending the principles of cloud infrastructure to distributed edge sites.
At the same time, a company called ZEDEDA was executing on the vision that we were building. In November 2019, I had the opportunity to join ZEDEDA and focus on my team’s mission full time. I liken ZEDEDA’s solution to “VMware Light ‘ – focused on edge nodes deployed outside of the traditional data center but still capable of running data center technologies like Linux, Virtual Machines, Docker, and Kubernetes.
In my role as VP of Ecosystem at ZEDEDA, I helped the team with core strategy and drove commercial partnerships. I was also responsible for working with the LF Edge community in support of our open source effort Project EVE within LF Edge. Working with the LF Edge Governing Board and Technical Steering Committee, we continued to build out the community’s operating structure, project base and thought leadership.
In 2020, “edge washing” was common and there was a lot of confusion. The “edge” was still mired by ambiguous terms like “Near Edge vs. Far Edge”, “Thin Edge vs. Thick Edge”, and so on. I helped lead the development of LF Edge’s taxonomy which broke the edge down into a continuum based on necessarily different technical tradeoffs. This garnered traction with the taxonomy among analysts because it was based on inherent technical tradeoffs instead of ambiguous terms.
A core theme of the LF Edge taxonomy was the benefits of extending cloud-native principles as far into the field as possible while recognizing the unique tradeoffs. Below Regional, Metro and On-Prem data centers, the LF Edge taxonomy carved out the “Smart Device Edge” for edge nodes that can run Linux but are deployed outside of the traditional data center. Companies like ZEDEDA and Balena are doing cloud-like orchestration for this class of edge device. We called the last tier in the continuum the “Constrained Device Edge” and highlighted that it was predominantly characterized by microcontroller-based devices running purpose-built, monolithic firmware.
In 2022, the community did a follow-on taxonomy white paper focused on Edge Management and Orchestration (MANO). The paper outlined four main MANO paradigms based on inherent tradeoffs that necessitate different underlying technologies and ecosystems despite sharing similar principles. It presented the evolution of technologies like virtual machines, Docker, and Kubernetes as they expanded further out into the field for Linux-capable edge nodes. However, at the most constrained edge tier characterized by devices that can’t run Linux, it assumed business as usual. Meanwhile, this tier equates to roughly two-thirds of the edge footprint by volume as is historically extremely underserved when it comes to development tools.
Atym is excited to announce that we’re seeding the code to start Project Ocre in LF Edge to address the constrained device edge. Ocre is a container runtime that leverages a combination of WebAssembly and Zephyr to enable app containerization and cloud-like orchestration for devices that can’t run Linux. It provides a hardware abstraction layer, supports secure boot, and supports isolated OCI-like app containers written in a choice of language (e.g. C, GoLang, Rust, etc.) that can be updated independently and can communicate with each other based on permissions.
The Ocre foundation enables a development experience similar to OCI containers and orchestrators like Kubernetes but is specifically designed for highly-constrained devices. With a footprint of only 256 kilobytes of memory, it’s up to 2,000x smaller than a Linux-based container runtime like Docker. Check out this blog from our CTO Stephen Berard for a deeper dive into the technology.
Ocre is an ideal foundation for OEMs and end users who are building and deploying software for highly constrained devices that are increasingly software-defined and powered by AI. These devices include sensors, controllers, IoT gateways, cameras, vehicles, drones, robots, and appliances that need to be continuously updated as AI models evolve and new security and regulatory challenges emerge. Ocre also enables a significant reduction in engineering friction and cost compared to traditional embedded development because code can be reused and different functions can be developed and maintained on their own lifecycles.
Speaking of Taylor’s 1989 album, we have this saying at Atym – “1990 called and wants its development tools back”. Our mission with Project Ocre is to revolutionize software development for embedded devices in the same way that VMware changed server application development starting in the early 2000’s, followed by technologies like Docker and Kubernetes. We’re excited to lead the charge with the Linux Foundation community to extend cloud development principles past the “Linux Barrier” to the final stop of the Magical Mystery Tour.
Stay tuned for more details on Project Ocre – we’ll be dropping code in LF Edge in May and making ties to other industry efforts focused on modernizing the embedded space. If building apps for tiny devices isn’t your cup of tea, take a look at LF Edge in general. There’s something for everyone spanning the overall edge continuum.
Bookmark the LF Edge website, follow us on LinkedIn, and join our mailing list to stay up-to-date on what’s happening.
