Edge Orchestration Services from Cloud for
Digital Transformation (ft. Microsoft & ZEDEDA)

Welcome to Advantech's InnoTalks, I'm your host, Matt Dentino. Today's topic is another in our series all about the digital transformation of our workspace. In our conversation today, we want to unpack edge orchestration services from the cloud. Now, I know that's a mouthful, so let me just simplify this just a bit. Edge orchestration combines edge cloud infrastructure, edge cloud communications service providers, and the edge-based applications all in a holistic or seamless offering. 

Unpacking Edge Orchestration: What Does it Mean?

[Host] - As businesses make that digital transformation, they're looking to move away from the centralized database centers to this edge for numbers of reasons that we're going to look at today. Things like bandwidth issues, latency, and then data security are all among some of the top reasons. As we unpack this topic, our guest today will specifically address questions for you and the big one, why it's important to you. 

We'll look at the industry trends that are our experts are seeing, investments that are being made by those major players in the space, and then the projected future growth. And why it's one of the most innovative movements for the decade. 

We'll look at topics of edge intelligence, deployment and infrastructure, as well as security. And we'll finish the discussion today with three real-world scenarios showing where implementation saved deployment time in one case, development dollars in another. And then in the third, provided great analytics while still remaining robustly secure. 

So, with me today to help unpack all of this, from Microsoft, Kumar Komanduri, IoT partner specialist. Pierre Lebas, the IoT solution architect. And from ZEDEDA, we have Jason Shepherd, he's vice president of ecosystems and a board member for the Linux Foundation's LF Edge organization. Gentlemen, welcome to you and thanks for being here with me today. Let me address this first question to Jason from ZEDEDA. 

Future of edge: the edge computing trends to watch out for

[Host] - Jason, we recently saw a Gartner Report on this topic that was called, Rise of the Edge. And they talked about projected numbers and the factors that are driving this move towards the edge continuum technology. Can you unpack for us a little bit from your experience, what is pushing that development and help us to start to answer that question, why is it important to us?

[Jason Shepherd, ZEDEDA] - Yeah, I mean, so if you look at the history of computing every 10 to 15 years or so, we see the pendulum shift between initially mainframe with very thin terminals connecting to it. And then, of course, the rise of the PC brought more power to the end user. And over the course of the years internet of course, drives new trends, mobile then brings some of that power back. Internet, of course, ultimately drives the cloud back and forth. 

But when you get to a point where there's so much data on networks as much as in a perfect world, we centralized everything in the cloud. It makes great sense, it simplifies things. You can do things without having to worry about the infrastructure below. All the great reasons why you would use cloud services, but it's not feasible to pump all of the data to the cloud. And if you have more and more connected systems, it's not feasible to make decisions on all of that data in a centralized location. 

So, edge is about-- this as you mentioned, this distribution of services across the spectrum, very important when people talk about edge. It does not mean the cloud is going away. By no means. It's going to continue to grow, it's just that it's going to be augmented more and more by compute on prem. We're going to see more and more workloads work on prem. 

But edge computing does not mean the cloud isn't involved. If you're doing edge computing without the cloud, you're basically doing legacy on prem apps. The cloud is often involved. Either it's processing additionally data that was pre-processed at the edge or collected at the edge, or maybe I'm training an AI model in the cloud using the unlimited resources of the cloud to train that model that then I deploy on the edge to act locally. 

There's a variety of things. Maybe the cloud is disconnected portion of the time but connected during certain times. There's a variety of different elements there. Edge, as you mentioned, it's bandwidth savings, latency, security, autonomy, in case I lose connectivity to the backend. Privacy, sometimes you'll strip private data locally, so that only a subset goes to the cloud. A variety of reasons. 

IoT is a key workload, more and more sensors drive more data that of course, drives that necessary shift and where data is processed. But so is edge AI. We see use cases around security and 5G is a big driver. A lot of people are like, oh, 5G is so fast you'll do everything in the cloud. Well, no, the connection locally is fast but you still have the same bottleneck upstream. So, you have to process more data locally. So, a variety of factors that are driving it today and we'll see that distribution increase over time. 

Extending Microsoft Azure Cloud Service to the Intelligent Edge

[Host] - Excellent, excellent. Thank you for that. Let me shift over to Pierre for this next question. And you can pick up a little bit on those trends, Pierre. But I'd also like to ask you as the solution architect for Microsoft, can you tell our audience how Microsoft is investing in this innovation and why in your experience, more and more customers are embracing IoT as a core strategy? 

[Pierre Lebas, Microsoft] - Thank you for having us. So indeed, from a market and a demand perspective, we are also seeing edge which is growing definitely from an IoT perspective but also from an edge computing by itself perspective to compute AI, and just to be capable to manage a growing and a more and more diverse workload on the edge. 

So, by the way, this is a way of edge computing is encompassed in a broader survey that we call hybrid, cloud or hybrid world, which is the continuum being capable to go from the tiniest device and to connect it in a proper and consistent manner up to an area to the cloud. So, to double down on what Jason said, it's not about edge or cloud, it's really how both are working together for the best value. 

And to point one single point of which is not just AI, but quite often devices there, which is a sustainability. There is no way from an energy consumption perspective and to project a world where you will transmit and push so much data from the real physical location to a backend, being cloud or any backend. So, we really recognize that the sustainable growth is well positioning, and this streamlined landing of additional workloads where they have to be best processed. 

And indeed, the reason; this connection tolerance, safety, latency, there are plenty of good reason why it makes sense, and consequently where customers are asking us what are the capabilities and what are the technological solutions that Microsoft can bring on the table to deliver this edge story. 

From another perspective, what Microsoft bring is a set of technology, typically software technology as you know Microsoft is mostly delivering software on cloud services. Or rarely hardware but generally mostly software part, which is going from the tiniest kind of device. So, what we call typically microcontrollers type of device, which can consume so little and which can be really very cost effective to larger type of device. Which can be a machine or which can be an edge, an IoT Edge Gateway which is the kind of device, which is agglomerating and starting to agglomerate to orchestrate locally. 

To what we call IoT Edge, which are more server class of devices because actually, when we think about IoT it's not always one cloud and many endpoints, but there is really some layers in between to start processing all the data for edge computing. And the good reason we have explained. And then out of this server there is of course Azure, or Hyperscale Cloud. But it is not exclusive. Generally, a good solution is the right balance, the right mix between all these different layers of edge to much the best business outcome. 

So, we have, just to name a few, Azure RTOS, which is going on a microcontroller Azure sphere, Azure IoT Edge, plenty of Azure things that you would love to learn more about. But all in one, the idea is to bring continuum on your offer and the most consistent way to manage that. Recognizing that we have set different set of technology to bring this consistency, some first party technology, and we have also a great partner ecosystem and good partners like ZEDEDA to deliver this layer, this consistency on the edge. 

ZEDEDA’s Open Orchestration Solution for the Distributed Edge

[Host] - Terrific information. Let me ask you this. So, we've looked at that structure, can you go a little bit deeper. Take me a little bit deeper on how ZEDEDA fits in the edge continuum using the Microsoft infrastructure. 

[Jason Shepherd, ZEDEDA] - Yeah, so obviously, Microsoft's got a really great portfolio as Pierre mentioned spanning, microcontrollers, with Sphere, all the way up through Azure stack, and you know of course up into the cloud. So, where ZEDEDA plays in this is there's a portion of the edge continuum and we'll get into more detail here that's kind of in between embedded IoT devices are very constrained microcontroller based devices, and traditional data center. 

And we call it the distributed edge. And what ZEDEDA does, we're a cloud-based orchestration solution. So remote management security of the infrastructure below the applications. We're not in the data path, we do not-- you do applications, we help you deliver those applications and manage the infrastructure. 

What we're basically doing at ZEDEDA is we're extending that, kind of the easy button of the cloud out of the data center as far into the continuum as you can go until you just have to go embedded. And we're doing that, so think of it as kind of like a take the principles of a VMware or an open shift, or any of these types of great companies that are doing things kind of in the data center. Extend those principles further out. 

We do this cloud orchestrator is what we sell, Z Cloud, that's what you get into and you manage applications by deploying them. Manage the hardware, get health of the infrastructure, et cetera, do all of that at scale. We leverage an operating system called EVE-OS from the Linux Foundation. So, vendor-neutral open source. 

EVE-OS is bare metal. You have that running on hardware, and in that that's what Z Cloud communicates with to orchestrate that hardware in any applications on top. EVE-OS is agnostic to hardware in terms of silicon. It is agnostic to the data stack, the application stack, and it has very deep security features, all the way down to the root of trust and the hardware. Is sort of like a TPM up through layers upon layer of things like measured boot, remote attestation. 

So, I know that the device is running when it's supposed to be running. I know that it is what it is. I can shut off individual ports remotely and kind of manage which applications can talk to what resources, whether they're processing or networks, and a bunch of extra features that you just don't get by running Linux on a box. EVE-OS's Linux plus an embedded hypervisor, and you know that actually makes it to where you can run VMs right next to containers. 

So, if you have a legacy, Windows applications, say SCADA system, HMI, Historian, and it could be a point of sale system in the retail sense. Any of those applications you can run on top while also running modern applications like Azure, IoT, edge, which we'll talk a bit more about. So full stack remote management rooted in the cloud. You have central visibility and control over all of your deployments. There's a built-in app marketplace that makes it easy to put in the applications that you want and just start clicking buttons to push stuff out. You don't need IT skills in the field to deploy things. 

So, it really is designed-- and when I say it's designed to bring that public cloud experience out to lighter edges, that means that we abstract the hardware complexity with EVE-OS. And so, all you're really thinking about as a developer or an admin is resources. I want this much compute, this much networking, this much storage for this app. I start deploying applications and I kind of build from there. 

And it's much easier to scale from that standpoint. So yeah, more on as we go but like that's the core principle that's as you said, it's cloud-based edge orchestration, which is a little bit of a mouthful. But yeah, it is about the easy button. 

Fit in Edge Continuum

[Host] - All that being said, there are a lot of parts and pieces in what we've just heard from Pierre and from Jason. So, let's talk to them a little bit about how these parts and pieces then come together. The implementation in concert, can you guys unpack that a little bit for us? 

[Jason Shepherd, ZEDEDA] - Yeah, I mean, I can start. So just so it's clear, so where we fit as I mentioned we're extending public cloud experience to the edge. But that requires a footprint for EVE-OS. So, our footprint for EVE-OS is roughly about 512 megs of memory and disk. Two CPU cores, there's always two copies of EVE running at a time for continuity. 

Below that, you have to go embed it. And this is where Pierre talked about the tools around Azure Sphere, and you know IoT plug-and-play and whatnot. That's kind of more lightweight embedded devices coming up. 

Above that, once you get into kind of the data center, there's a lot of great tools for the data center. What's different in the middle with EVE-OS and what we're doing from ZEDEDA standpoint is data center. You can assume the servers are locked up. You can assume that it has a well-defined network perimeter. We assume someone could walk up and touch that box. We assume that it's on a network that you don't own, so you need to be able to do segmentation and secure things. 

We assume you're going to lose connection at times whereas data center stuff assumes that you've got a constant connection between a controller and a server, you're usually on fiber. And so, the operating model is flipped. So, we're in that in-between from a ZEDEDA standpoint of the broader portfolio, where you can run cloud native principles. You can support virtual machines, but you have to assume certain things because you're not in a physically secure data center or a traditional data center. There's an in-between spot that we are very focused on. And Pierre, I'll let you kind of add on top for the Microsoft side. 

[Pierre Lebas, Microsoft] - Yeah, I like what you said about all the diverse kind of workloads and hardware that have to be managed. Because a lot is about it, is about managing the legacy, what is already in the field or running in operation, what is coming with some new innovation, and, for example, the Azure IoT Edge is a good enabler for that, and all the different kind of hardware, which is underneath and how to make it all consistent and cloud manageable. 

Because if you think about it 5 to 10 years ago, a lot of these things-- of this legacy was already existing, but it's just that it was a managed very bespoke on purpose for different use cases, but very siloed and isolated. And what is really the vision here is to bring a consistency, and really a maintainability which then will allow to scale and to unlock more larger deployment, and more business outcomes. 

So, I guess that's where we will probably speak just after a while. A bit more on details on how all these things put together will bring best capability and scalability. 

Joint Solution among Advantech, Microsoft and SEDEDA

[Host] - All right, excellent, excellent information and we're diving deep here. So, let's next talk about how and where ZEDEDA and Microsoft intersect in the deployment. And then how that intersection helps to scale. So, let's jump into that part of the conversation. 

[Jason Shepherd, ZEDEDA] - Yes, I mean, I'll start. So, we've talked about the portfolio on both sides, Microsoft obviously has a very broad portfolio covering the spectrum. We're focused on that middle section, which also happens to be exactly where Advantech play. Is in kind of this-- the spread of ruggedized hardware from gateways up through servers, and various different processing types. And so very, very close alignment, which is also why we see Advantech as a very strategic partner. 

What we've done specifically with Azure IoT working with the product teams is we've integrated ZEDEDA with Azure IoT in the context of Azure IoT Edge. So again, if you're using Azure IoT with very lightweight constrained devices or you want to doing microcontrollers that would be sphere and whatnot, like we've talked about or plug and play. 

But if you want to run software-defined multiple applications on a box, have you head headroom for future applications, I need to run virtual machines next to Azure IoT Edge for like legacy stuff or other types of things. Anything you composite. What we've done as an integration where ZEDEDA serves as the infrastructure plane, EVE-OS is that bare metal operating system. Our cloud is how you manage that hardware at scale. Could be any number of different types of hardware in a broader fleet. 

Azure IoT, of course manages in-band modules deployed in Azure IoT Edge. All the tools that you would use from Azure IoT they're all there, and you would use them as you normally would. Where we would come in is we make it easy to deploy Azure IoT onto a box, manage the whole lifecycle of that box out in the field at scale. Secure it down to silicon with things you mentioned before, around our security models zero trust security. And we've done direct integration between Azure, the Azure console, and ZEDEDA console, so you can see what modules are deployed from Azure on the box in the ZEDEDA console. 

And that separation of concerns between infrastructure plan and application plane, I mean, this is IT 101. Like I always say when's the last time your ERP system manage your PCs? Don't do that. You separate these out. The person that's doing DevOps, data science, that line of business, looking at the outcomes, you're going to live in Azure IoT. You're going to live in that suite of services. 

If you're the infrastructure admin, the network person, you may be out in the field deploying things physically, you would leverage the ZEDEDA cloud for that infrastructure management. So, it's a very complimentary set of options that keep separation of concerns between the different needs, and also provides more flexibility in terms of what you can deploy together. 

Run Well-Architected IoT Workloads to Create Business Value

[Host] - OK, excellent, excellent information. You guys are a wealth of knowledge. I'm learning as we go. Let me address this next question now to Kumar, bring him in to the discussion. So, we've got a lot of information on the technology and on the infrastructure, can you talk to us a little bit about how-- what are best practices to pull that all together? How are we and what are we using to get this up and running with confidence and focus? 

[Kumar Komanduri, Microsoft] - So, thank you, Jason and Pierre, I think those are great points. I would start with-- I mean, maybe looking at the a little bit and start with IoT in the cloud and IoT on the edge. Now, Jason has referred to how the edge ecosystem is proliferating. So, if you look at IoT in the cloud, is a symmetry that's going to occur in the future. There's an IoT in the cloud, and the workloads on the cloud are going to transition into edge. 

Now, part of the reason is edge has a greater requirements such as low latency, tight control loops for a real-time response, and you also have protocol translation and data normalization. And there's also concern of privacy and data protection of IIP. So, there is going to be a heavier symmetry that's going to lie on the edge, and which is where the orchestration and the device management, all these key principles become imperative for a successful implementation. 

Microsoft lays out this solution from secure from silicon up, and the strategy is both for greenfield and brownfield ecosystem. So first, let me chime into the eight attributes that are key for a successful IoT implementation that are rolled into the five rough pillars. 

So, the first one is scale. I mean, what are your growth strategies that allow you to scale more seamlessly? For example, one such strategy could be a grouping device into a larger geographically distributed deployment. The other one, the second one is device management. How do you manage device at scale? And not only considering how to manage devices themselves, but also how to engage the data they are sending. 

The third part is the big data management. I know this is more of on the cloud side, the IoT applications capture, store, and analyze, and act on huge volume. But I think how you choose to store, retain, and organize your data sets a foundation in terms of long term implications. 

The fourth one is the analytics, insights, and actions. So when people think of IoT insights, they only think about real-time device health and performance monitoring. However, evaluating device data on aggregate and over time can provide priceless insights. The next one is the high availability in DR. I mean, for a major critical applications and data, you need to build in resiliency and so, that has to be baked in from the point of deployment. 

So, you need to architect your solution with fault tolerance and disaster recovery in mind from day one. Again, the next part is the security and compliance. Security is an overarching strategy for maintaining security and ensuring regulatory compliance. And I can definitely touch upon the seven properties of highly secure devices. I think customers who are deploying IoT whether it's the brownfield or greenfield, they need to consider from silicon and start from hardware of trust. 

And there's a seven properties of securely-- highly secured devices. There's a white paper on that, I would definitely as partners and customers to review the document. 

8 Attributes of Successful IoT Solutions

The next one is the managing IoT solution with DevOps. I think the governance of planning, testing, and deploying, documenting, and managing different components of your solution is a key. And in a coordinated way, and I think that's where the orchestration comes into play. And the last one, the eighth one is the understanding total cost of ownership. 

Understanding how you spend to run and operate all your services involved. So again, Microsoft Partner solutions on hardware, software, are reducing their overhead. You can-- there are different package solutions and offerings that can to get you to speed to market. 

So, these are all eight attributes of the Azure IoT, successful IoT implementation are rolled into the five pillars of IoT well-architected framework. So, the five pillars of IoT well-architected framework are reliability, cost optimization, operational excellence, performance efficiency, and security. 

Microsoft published a white paper on the detailed documentation, including assessment on the IoT well-architected framework. I would encourage the customers and partners to use that as a framework for you to build your scalable IoT deployment. As we move on from-- again, as I mentioned, I been going to talk to customers I usually refer to 758. That is 7 principles of highly secure devices, 5 pillars of the IoT well-architected framework, that should encompass your 8 attributes of successful IoT implementation. 

So, starting with 7 properties of highly secure devices, I think you need to have a device catalog where as you can see, there are different partner solutions that where for example, Advantech has the industry certified devices and gateways that can run Windows IoT and IoT edge platforms. 

And there are different programs under Azure device catalog. So, the end goal is to get the customers to market in a secured and in a faster way. And these are your plug-and-play certified devices. They start from sensors, your Azure Sphere, and your gateway class devices that Advantech rolls in with hardware to be the TPM modules that secure hardware. 

And there is a solution aggregator, there is a solution packet solution platform. And there's also in future there's going to be edge secure core gateways where there's a program that's going to be implemented where there's a higher security framework that is deployed right from the-- right from the hardware. 

I think I will highly encourage customers to review the IoT well-architected framework, the 7 properties of highly secure devices to ensure a successful IoT implementation. 

[Host] - That we'd finished today by providing three real-world scenarios where all the stuff that we've talked about has actually been deployed successfully. And so, I want to ask Jason if you would walk us through, give us that final rundown of these three scenarios. 

Customer Examples

[Jason Shepherd, ZEDEDA] - Yeah, so very common pattern. We strategically did that integration with Azure IoT edge because a lot of customers are coming to us using Azure IoT. But again, it's kind of that middle ground between constrained devices in the data center. And so, very common pattern. A customer decides hey, I want to use Azure IoT and all the services associated with it, et cetera. 

And we were actually in the Microsoft marketplace and people find us because deploying Azure IoT edge there's a manual process associated with it that we automate. And we also give that visibility I mentioned between the different consoles, kind of tie everything together because there's a deeper integration. 

And so, a couple of customer examples. So BOBST-- I mean, BOBST and PEOPLEFLO are public case studies. And then we have one from a very well-known manufacturer that's not but we'll talk through. 

BOBST was deep into Azure IoT. You've seen the value but had to be able to make a decision like, do I go figure out how I build out the infrastructure below it, or do I partner with somebody or work with a technology provider? From the time that they found us through the Azure marketplace at the time we hit production was five weeks. 

And this is the value of having kind of productize infrastructure. You make it easy to deploy Azure IoT edge, we were now scaling into the thousands of deployments and this is key. Is you need this as foundation. So, to build and really scale things out, people flow very similar. They're doing-- so BOBST is a big machine builder. They make connected machines. They sell these machines to their customers and deploy new types of services running compute inside of those machines. 

One of the things there is when you sell a machine as a machine builder to a customer, you don't own the network that it's going into. So, you need to be able to work in reverse through that network to your own networks and services. So that helps there. People flow using Azure IoT to do AI to drive pump efficiency for these big industrial pumps. Same thing, I needed help deploying Azure IoT edge and simplifying-- kind of managing that infrastructure below. 

The manufacturer example, this is where when you go direct to an end user whereas the first two examples are machine builders that then send-- sell their services to their customers. 

When going direct to an end user, here it's important and it kind of comes on the technology providers to balance the needs between the operations group, OT, and the traditional IT groups. So, in this case, it was about the operations group wanting to pull data out to drive analytics with Azure IoT. And the question is, well, can we put a box on the network? And IT's like, nope, you can't put that on my network. 

So, we went through a security audit with IT and because of the security models from a zero trust perspective with ZEDEDA, this notion of built-in network segmentation with distributed firewall, et cetera, et cetera, it became to where it's good for both sides to build and deploy this infrastructure on premise. 

And this one actually demonstrates the composite nature. So, this is using Azure IoT edge as a runtime connected to Azure IoT, as well as Defender, which is formerly CyberX IoT threat analytics software running concurrently alongside Azure IoT. And it also then you have headroom of course to deploy additional applications over time. 

Several examples of this pattern that we see very often where people want to use the power of Azure IoT but also want that kind of easy button infrastructure layer below to give them maximum flexibility. 

[Host] - Very good, excellent. I appreciate your tying that all together for us and bring in those real-world scenarios to it. It makes perfect sense. I want to-- we're just about at the end of our time. Pierre, I wanted to see if you had any last comments. I know that there may have been something there that you wanted to add. 

[Pierre Lebas, Microsoft] - Just maybe just closing that we are remembering that Microsoft is building great platform, good product and is really leveraging on the capabilities of a strong ecosystem like ZEDEDA is a part of to bring the-- to match the real customer and market demand. And in this case, related to edge computing which is clearly growing and that we want to address together. 

[Host] - Excellent. Thank you very much for that. Last words go to you, Kumar. Anything you'd like to add? 

[Kumar Komanduri, Microsoft] - Sure, yeah. Now, I think I would definitely recommend customers and partners to go to the device catalog. I think that's to speed to market deployment. And there are Azure certified devices that are IoT plug-and-play devices. There are different categories and that starts from sensors, microcontroller, like Azure Sphere, that are highly secure to Advantech's device or Gateway portfolio that can run both IoT Edge and Windows IoT, and Linux flavors depending on to simplify the workloads. 

And I know we are excited for all the edge market and for customers bringing in the intelligence to the edge. And there's a broad range of solutions and hardware catalog that's available to get them started right away. 

[Host] - And with that, we've come to an end. I want to thank my guests for sharing their time and their expertise with us. Kumar and Pierre from Microsoft, and Jason from ZEDEDA. Gentlemen, your information and insights have been invaluable. I want to thank you. 

Edge Orchestration – 5 Takeaways You Can’t Miss

Just to recap, we've gone through the whole edge continuum from structure, to purpose, to actual implementation. And we looked at different reasons to Jason's point, we're trying to find that easy button where we can solve the problems of bandwidth, latency, and yet still, have robust security in our distributed cloud-based applications all in a wrapped in a continuous or contiguous infrastructure that follows Microsoft's well architected format and function. 

And we talked a little bit about how ZEDEDA implements and also leverages that infrastructure, and that back end. And we heard three real-world scenarios where that implementation was saving in development time, implementation dollars, and also in security. 

So, with that and to our audience, I want to thank you for joining us here with Advantech's InnoTalk. And as always, if you have questions about anything that you've heard today, you can reach out to us at www.advantech.com, as well as iiot.tech@advantech.com. We look forward to seeing you online and answering any and all of your questions that today's discussion may have stirred up. For Advantech, I'm Matt Dentino and I hope to see you here again on our next InnoTalk. Until then, so long.