Back in February, I was invited to be a delegate for the first Edge Field Day (EFD1) from Tech Field Day. I’m going to lump my impressions from that event together and include some sad news from the Edge world that happened recently regarding the Intel NUC. All is not doom and gloom in NUC world though, as I’ll explain later.
Edge is a vague term that has a lot of different definitions. Some treat Edge as the Edge of the Network; others focus on bringing Compute to the Edge. Prior to EFD1, I didn’t have a full grasp myself on what Edge truly meant myself. My focus was on how Edge is being used to reduce latency with Private 5G. EFD1 presentations were more focus on the Compute at the Edge. Oh and there was one more thing.. NUCs everywhere!
Private Cellular and Edge
The wireless world is bringing Edge to the forefront with 5G use cases. My foray into the Edge world began with our Private LTE Network. At the beginning of our design, we quickly realized that the Private LTE Data Plane separation from the rest of the EPC Core can be useful. Celona, Pollen Mobile, and others are using NUCs for this very purpose.
In Private Cellular, you can have a central core but separate where the End User Equipment (UE) connects to the Internet. Years ago when I first started to research cellular technologies, I read that it was possible on a carrier network to have your Data Plane traffic terminate at one data center on the West Coast, somehow travel across country without your device updating to a closer data center. The carriers typically have data centers all across the country that drop the user data off to reduce latency and other reasons. This can be useful from a Private Cellular perspective!
For the purposes of Edge, it is bringing the compute or connectivity closer to the device to improve performance and reduce latency. That is how some of the Ultra Reliable Low Latency Communication (URLLC) is achieved in 5G. David Calisch from Celona shared a graph about Latency that I found interesting. Although, URLLC requires 5G NR Stand Alone (SA) to truly benefit and most current 5G implementations are Non Stand Alone (NSA) at the moment. I’ll dig into these details in a later post because that deserves a more in depth discussion.
Edge Field Day 1
At Edge Field Day 1, we didn’t have any vendors show off how the Edge is involved with Cellular although we did have some interesting Network Edge products. Almost everyone was involved with or partners with one company that in some ways is building the Edge use case in their product field. There is one item that came up a lot throughout all the presentations.. Intel NUCs.
Scale Computing
I was a remote delegate, but the presentations mostly were located in the offices of Scale Computing. An interesting thing happened a few days before EFD1. One of my VARs that I work with, setup a demo with Scale for me. So, I wasn’t completely blind to their product line.
When it comes to Edge, Scale is a big player. Their Hyperconvergence technology allows organization to bring compute closer to where the data is being collected. Using Docker and Kubernetes or other baremetal or virtualization technologies, they can provide a hyperconverged system that processes data at the edge. Docker and Kubernetes allows organizations to implement edge on regular COTS computers like the Intel NUC.
Where Scale Networking really shines is the management of edge devices, the hardware. Scale partners with other companies like Avassa and Zededa to handle the application management side.
I’m not a server guy, although I run a few small data centers, but I have worked with industries where Scale and edge are being used extensively. I was raised running a successful fast food restaurant and small motel. My first job out of college was working for a Restaurant Point of Sale (POS) company. I watched those industries move from on site client-server based Point of Sale systems to iPads and Cloud Computing. In fact, I installed the first iPad Point of Sale terminals at a Restaurant in the state of Utah, two weeks after the iPad was released. We have come full circle were restaurants want compute locally for some systems while relying on the cloud and its benefits for others.
One of my fellow delegates was Brian Chambers from Chick-Fil-A. Right before EFD1, Brian wrote an article about how Chick-Fil-A uses Intel NUCs to create a common compute architecture. Franchisees are given several Intel NUCs to run all the compute needed locally while giving corporate power to modify and manage the systems. They are running Kubernetes with Docker.
Because of my background, I was fascinated with the system they have built at Chick-Fil-A. Right out of college, I dreamt of a world where we built these POS systems that communicated with a central data center and provided central control. We had a couple clients that could have built something similar on a smaller scale than Chick-Fil-A and it would have been super cool. The technology just wasn’t available yet. It’s exciting to see those ideas being used today.
When rolling out Edge applications and hardware such as the Intel NUC to sites to bring compute closer to the clients on a Private 5G network, Scale is one of the leaning vendors. If I had a use case for deploying a large number of edge devices with private 5G, I would absolutely be looking at Scale for management of my hardware.
Avassa
A big partner of Scale is a company called Avassa that also presented at EFD1 in Scale’s offices. They basically built their technology on top of Scale’s architecture. Scale is for managing the hardware, Avassa handles the software. Avassa uses Scale as the backbone. They provide management of the Kubernetes and Docker containers located at Edge sites. This will be useful as we role out sites that are using Private 5G.
Zededa
Like Avassa, Zededa is a partner of Scale focused on the edge application side. Zededa has an in depth page that discusses where their products shine. As they say on that overview page, their focus “provides an edge orchestration and management solution engineered to deliver applications and workloads to edge devices.” Their flagship is a product called EVE that is a baremetal virtualization engine. Like Avassa their main focus is on providing management of software at the edge.
Although working in similar ways, Avassa and Zedada have a partnership to solve particular problems together. For anyone looking to build some edge use cases on a private cellular network, I would dig deep into what all three of these vendors have to offer.
Mako Networks
Mako Networks was something different at EFD1. A couple weeks before EFD1, we had a pre-show presentation with Mako and many of us were left wondering why Mako existed. After EFD1, we understood their niche.
Mako Networks focuses on Point of Sale networks with networking hardware that is hardened for PCI Compliant sites. Instead of buying networking gear from any vendor, Mako helps a restaurant or retail site ensure their network is more secure because it is controlled end to end.
They do not operate in the Private Cellular space nor rely on the trust Intel NUC, but they do have wifi and switching hardware. Their focus is just on the Point of Sale market that requires PCI Compliance and their product pitches makes sense toward that niche.
OpenGear
Finally, we got to hear from OpenGear about their approach to Edge. I was excited to hear about them because of their ties to Digi, a company involved in the Private Cellular world. While OpenGear equipment could be tied to PCNs, it makes more sense to use a carrier cellular network in most cases.
Their products are to provide management of networking equipment (think console port to routers, switches, etc.) at the edge site. Unless a site has a completely separate network for a Private Cellular, it is recommended to use Carrier Networks since they are out of band of the devices you are managing with OpenGear. I was a little let down, but really liked their product line. I also was excited to find out they were a company based in my home state!
Intel NUCs
Prior to EFD1, I didn’t know that Intel NUCs were used so much in Edge sites. I had been looking at possibly building an Edge use case into a site near the 4G/5G radio edge but didn’t think the industry was doing the same. At my family businesses, I’m currently using NUCs for POS compute at the edge as well as management of a LORAWAN system for tracking refrigeration temperatures and humidity. I didn’t think of those two use cases being true edge. That’s the thing about edge is the definition is a little hazy, just like how cloud was misunderstood for a long time.
At my family’s businesses, I’ve been using Intel NUCs for years. When NUCs were first announced, we explored using them at the Restaurant POS company I worked at. We had clients using them as client-server on Food Trucks. The Apple iPad and 4G Networks weren’t as mainstream as they soon became, although those soon took over.
Cloud and iPads have changed that industry for the better and Intel NUCs are used more for other systems at this point, like how Chick-Fil-A is using them.
When news came out a few weeks ago, I was saddened to hear that the end of an era is upon us. Intel has decided that the Intel NUC is no longer worth pursuing as a business. Obviously, someone will pick up where Intel is leaving behind. There is just something special when it comes to an Intel branded NUC instead of the plethora of other vendors that have copied the market.
Future of Edge Without Intel NUCs
So what is the future with Intel officially sunsetting their version of the NUC? It will probably be status quo, just not relying on the trusty old Intel NUC. There are enough vendors that have copied Intel that the NUC market idea isn’t going anywhere. They just may be more AMD based devices that are going to actually make it into the market. Simply NUC and others have announced they are taking over the NUC line from Intel. The market is migrating from Intel running the show to a partnership.
The Edge is migrating and still being defined. I’ve explored devices from Cradlepoint that allow you to run very limited Docker containers on their CPE routers. Those are ARM based. The Raspberry Pi brings a whole option for some low requirements. The usefulness of NUCs is proven. Just who will take over at this point? Who will replace the trusty NUC? We will just have to wait and see how this shakes out.
Edge Is Here To Stay
After listening to the vision from the vendors at EFD1, I can see that Edge in whatever version you attach the definition to is here to stay. It’s just an evolution of the cloud model; another tool in our belt. I’m so glad that someone is thinking about these intersections. So, that when Edge and Private 5G start coming together we are not having to design these system from scratch. Bring on the Edge and Private 5G use cases!
