Kubernetes IoT Edge Working Group, 12 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Edge Computing Is Hot. Find Out the Business Value From Three Experts

Description

Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon Europe 2023 in Amsterdam, The Netherlands from April 17-21. Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

Edge Computing Is Hot. Find Out the Business Value From Three Experts - Larry Carvalho, RobustCloud LLC; Stu Miniman, Red Hat; Marilyn Basanta, VMware; Muneyb Minhazuddin, Intel

Edge computing is drawing increased investments by organizations gaining efficiencies by taking advantage of automating physical activities. While consumer applications with connected homes have high awareness, significant business benefits are gained from using edge technology in commercial applications like the factory floor, mining operations, etc. Simulation of a digital twin of a complete factory or warehouse by embedding technology into devices can assist in making better decisions without large investments. Technologies used in edge solutions are AI/ML, 5G, IoT, and cloud native architectures.

This panel discussion will discuss how cloud native tools provide business benefits to applying edge technology for physical process efficiencies. Larry Carvalho, Principal Consultant at RobustCloud LLC, will moderate this session. Stu Miniman from Red Hat will highlight how a cruise line operator used a Kubernetes-powered edge solution to deliver an excellent digital experience to ship guests in disconnected environments. Marilyn Basanta will discuss how a large ground shipping service was able to add a modern application platform to their distribution centers, ensuring higher employee productivity and business continuity during natural events. Muneyb Minhazuddin will talk about how a manufacturer modernizes applications on the factory floor as they converge IT and OT to create a dynamic and agile production line in a factory shop by software defining constrained devices.

Join this session to learn more.

A

Okay, I hope you're ready to come here ready and for some Edge Computing stuff that we're going to talk about Edge Computing is hot. What this session is to talk about the business value of edge Computing, what it is not is to what is Edge Computing we're not going to talk about that. If you want to go more in depth, there's a lot of Linux Foundation stuff that you can download.

A

My name is Larry Carvalho I'm, an independent Analyst at robust cloud and with me I, have you know three panelists I would like them to introduce themselves starting with Marilyn.

B

I get to go first, hi everyone, my name is Marilyn basanta I work at VMware and run all the product management for our Edge Computing platform. Still.

C

Hi I'm Stu miniman I've been coming to the kubecon since the early days and back at the openstack shows before when we were starting to talk about containers, I joined Red Hat two years ago, director of Market insights, so part of the openshift team. Before that I'd spent a decade as a analyst and host of the cube when.

D

You hey, my name is I'm the chief marketing officer for networking and Edge business at Intel um I've, just been there 45 days before that, I was at VMware building the edge Computing platform as the general manager.

A

Okay, so what we put together over here for you is to talk about different use cases of edge I mean Edge, is running all over the place. When you think about agriculture, when you think about mining, you think about Healthcare, we've got a lot of edge Solutions. What is the business value that you can drive? Maybe in manufacturing can you you know improve maintenance. You know throughput improve quality. Those are the things that we need to think about when we, when we bring up Edge.

A

So with that what I'm going to start for the first is with munim to talk about one of his use. Cases spend some time on that, then we'll have the rest of the other. Two panelists talk about their use cases, and then we will go to you know more of a discussion. Q, a uh you know, within the panelists and outside the panelists, about that. Okay. So with that money, I'll bring up your slide and if you can't.

D

Yeah um I think we decided to kind of showcase what we did for um you know smart cities and you know, build out a roadside unit. I know we said we won't talk about what a definition definition of edge is. You know, Edge is an interesting phenomena, that's well.

D

It has been there here for a long time, but um what we see is you know workloads being built at the edge more more often now than in the past so um and, like Larry just said they manifest themselves in different verticals because they saw it all being particular problems. So I.

C

Think all of us.

D

Have a vertical use case scenario: that's happening in this instance, um and you know there's a lot of convergence of Technology happening there too. In this instance, we've converged 5G connectivity to AI. uh You know inferencing at the edge, because we feel inferencing at the edge is a major driver, a lot of new data being generated. How do you infer that a lot of that is trigger based there's a lot of data coming at you? Storing that data is not a lot of value, but you know inferencing. What outcome you want to come out of.

D

It is what the value is in. So how do you actually do inferencing? So what we built here with capgemini as a partner and we're testing this in cities of London and and here in Italy, and they are roadside units which are connected through 5G and they're, spread around the cities and they're doing centralized traffic Management on you know multimodal transport. So you know your road, your Railways, your Port, all of that together and coordinate it.

D

So there is a level of a control plane which manages the Central City, uh um think of it as a city level. Operations and then there is individual Edge units and each unit has 5G connectivity. It has a small Edge compute platform. It has, uh you know Vision computer vision all built into it, so we're doing um inferencing of those applications and I'll follow up with you know in in further kind of questions. You know certain scenarios, but you know, traffic management is a start.

D

um We're also doing traffic management, for you know we're testing with emergency services, so I think that's the important part. So this is ambulance fire. um You know police I, you know those are kind of the the high stressed situations and uh the application Level compute that's happening is just inferencing. Looking at traffic patterns. Looking at you know, incidents looking at all of that and centralizing it and uh you know, we've used. uh You know our 5G software stack as well.

D

As our you know, AI inferencing stack and run it on a really thin Edge uh kind of platform which is built out so um amazing can I take. We have multiple cities around the world, I willing to kind of jump on this, so yep.

A

So, as you can see on this slide, you have you have all of these Open Source Products that have been used. You see what Intel products are used and what I find fascinating about. Traffic management is all the edge devices that you need everywhere from from the from the roads to you know to the utilities. When you have it, you know traffic signals Etc that all have to work in cohesion with each other. So with that we'll go to our you know.

A

Next with Stu, you know who's going to talk about what Edge Computing is in defense and security. Yeah.

C

Thanks Larry so uh yeah the use case we're talking about here. This is actually the central I.T Department uh for the Israeli defense fund. A group called mamram and you know it's funny. I I think back many. The the first conference I went to that was like really talking a lot about containers. Was you know back, uh you know, IDF the you know which we used to joke, it's the Israeli defense fund, or is it the you know, Intel developer Forum?

C

My daughter had a backpack from that one and people would get confused um so um but uh the Israeli defense fund that when we talked about containers in like the earliest days- and it just reminder some of the basic things, what can I do? I can spin things up faster. I can I can just accelerate my applications uh because the the unit of it is much closer to that application, and so, as it shows here on the slide, you know things that used to you know.

C

Oh, we used to do our build and we used to do it with VMS and it took us weeks, and you know now we can take that down to hours and the other thing you know, I think a common theme you're going to hear from all of us AI ml or great workloads here, because again, if it's I need to be able to do I, do my processing back at some Central I.T in the cloud or can I do something at the edge.

C

So I heard a great line recently that at the edge, what we really need to become really efficient is killing the data at that location, because um we we all know how there's that explosion of data at the edge but I, don't really need all of that. So in this case there's many times they want to be able to process it there. You know: we've got kubernetes running up on the International Space Station.

C

Obviously, if I had to think about every time, I process, something you know sending it back down to the earth, that could be a challenge in a lot of cases. So again, some some really interesting things. One of the open source projects uh that we we list on here, it's uh kubeflow is the basis for the the open data Hub.

C

So a whole lot of data science, tooling, that's available that this customers and many others are being able to take advantage of so yeah I think just accelerating how helping their developers we did a lot of training for them to get more people up and running, because there's that dichotomy, there's new skill sets that we need to learn, but at the edge. Typically, you don't have the people resources. So that's why we have to have a lot more automation. We we want to be able to.

C

You, know it's not taking people out, but it's having people in the right location to be able to take care of what we need and typically at the edge you're not going to have that highly trained skill set to be able to take care of it and- and the last point on this, of course, that you know General message. You'll hear from Red Hat is consistency everywhere, so Linux lives everywhere. Openshift can live everywhere. That Linux can so this environment was based on what's in a single node openshift.

C

So if you know just a kubernetes architecture, normally I need, you know how many compute nodes and worker nodes and how many systems that is well single node openshift allowed us to boil that down to a single machine with both the control plane and the worker nodes, which helps, of course, if it needs to be disconnected. So this can work in you know semi-connected or disconnected environment for time, and when things get reconnected they can they can.

C

You know, share the data that they need, so I think we'll have some more questions to dig into it later, but you know that that's the overall right thanks.

A

Thanks so one of the things you're seeing over here is the whole mlai aspect of edge. What what I have seen as Edge has been progressing at three different areas, that there is some Innovation that is helping Edge. There are a lot more, but the three that I see obviously is mlai on the edge where you can use it for image. Recognition and several other aspects of uh you know what what can be done on the edge. The second part of it is 5G and networking. You know the speed that you need to.

A

You know, connect with these devices and obviously all the security that that goes on with it, and the third area is more Hardware acceleration at the edge where you have specialized chips or made for purpose chips that are built for the edge and I think these three aspects of Technology evolution is really improving the evolution. You know off edge in the market and in this case you're seeing one of them, but obviously behind the scenes. It's using the other three aspects, the other two aspects of edge.

A

um So with that again, you see the Open Source Products. What what red hat products are being used in this solution and with that we'll go to to Maryland, to talk about the VMware story about the global food right there yeah.

B

Hi everyone so um similar story, so we've got smart cities, we've got defense and so I'll talk a little bit about manufacturing or processing food processing. So um this is a customer. It's a food processing customer in the U.S and they've got multiple different processing plans, and so what what they've been able to do for each different processing line? They've deployed our Edge compute offering and they're doing they've got cameras all over the line and they're basically doing inferencing at the edge as well.

B

It's interesting I didn't think we were all going to talk about inferencing, but we they're doing inferencing on the line, meaning that they've got the different workers physically taking the meat off of the bone and what they're doing is they're using the cameras and the inferencing technology.

B

So after the worker has done their job and put the they separated the meat from the bone, they put the bones back on the line, and then the inferencing is actually analyzing that in real time and then as that bone comes down the line, if they realize that maybe the worker didn't get all the meat off the bone or there's something wrong with it. They can kick it back onto the line to come back to be processed again, and so it kind of does a few different things.

B

One is, it becomes more business efficient for the company and then second, they become more sustainable, as we know how how impactful it is, for you know, to grow a cow and to grow animals for our food processing. It's very important for us to be able to be more sustainable. The nice thing a few advantages here is that this particular use case for them has actually completely paid for itself for them to enable than other use cases. They can do then do predictive maintenance because they've got they can connect the sensors onto our platform.

B

They can also, then do more well. I mean what they're doing also is close with automation, to be able to change the different arms to kick that bone back onto the line, and they can also then do training with the workers, because they're monitoring, as the workers are working they've got it for safety if something were to happen with the workers as they're doing their job. So it's just it.

B

It's having that edge competing platform at each of these processing lines, besides the economic advantages and opens them up to be able to do all the other different use case. Cases of connecting so many different discrete systems together. I think the super interesting part about manufacturing as well is that manufacturing is bringing together there's still a lot of use.

B

Cases happening physically, so when usually other businesses and previously have done inferencing at the edge they'll do that on the physical box with the cameras, it's a completely separate system from anything else that happened on the factory. So that's all the hardware for all that they have so with our Edge compute platform. We can really bring that together and make it more efficient to run their existing workloads. In a lot of cases are still VMS, as well as help them modernize and I'll also say with a lot of these existing workloads.

B

They can completely cut out the VM stage and get directly to kubernetes, which is another great part of it. So it's just interesting all of the different use cases to come together to to solve for this. um First of all, for all of the different use cases in manufacturing.

A

Great, thank you so I'm going to follow up with first of all here you see another example of uh you know both image recognition as well as AIML, which are connected with each other. This is also related to the amount of data that is processed now at the end and how much of data growth is at The, Edge versus the cloud, and- and that's that's expected to be in many more. We have you know multiple x times of what is in the cloud is, would be on the edge in the future as by 2025.

A

There are several numbers being run out there and I would like some of the panelists to talk about that later, um but uh Maryland. If we think about this use case that you brought up how much of waste did the company you know reduce, and one of the things why companies get into these kinds of things is, is the recognition by their customers? You know folks are looking at ESG. They want to see. You know how sustainable you are.

A

So did this company get any recognition from from the um generally the consumers is as to how they were doing it, and you know what what would you talk about these two aspects of the solution? Yeah.

B

Only of course, could Market. That is an advantage to them. I, don't think they've yet received any public recognition, but I mean the overall for them is it's important like I said statement to make um and I do think? Is it's becoming much more critical around the world, especially with everything that's going on, and so it's um it's been a good thing for them. What I wanted to say for them as well is they've. You know they think you wanted to know about, maybe particularly um how much savings this is or what did.

A

You asked sorry.

B

Thank you. It triggered my memory. um Yeah they've definitely been able to see significant reductions. I can't share the exact numbers, but it's you know double digit reduction as well. As um you know, for the cost savings you know. I can't share the exact number, but this definitely has been um it's properly efficient use case that pays for itself.

A

Got it got it so for Stu, you know talking about the Israeli Defense Force, you know what what do you see over there? You.

C

A

What would the new mlai processes that they did? What what it value did they get? You know without obviously giving up any state Secrets over there. Yeah.

C

One of those I could tell you, but uh somebody would probably kill us all, um but no um yeah, it's really important. It's interesting. We often talk about Dynamics. Oh it's is it the cloud or is it the edge um this unit? They actually call their Edge thing, they're called cloudlets um and that if you really think about just architecturally, um it's all connected. So you know the the typical model that we see is AI often I. Do my training in the cloud um and I do my processing at the edge.

C

So you know, we've worked with a lot of the auto manufacturers. It's like! Well, you know if you bought a Tesla or some other car there, um it's not going to create the model in your car, but you want the latest software push to you um and you need that processed immediately, so that you know you're not going back to a central location.

C

So it's same thing in this force is it has autonomy, but it is still tied into in feeding information back to Central I.T, and that's really just a central tenet of you know, constantly relearning and uh iterating on what they're doing from the field right.

A

One one question: I want to ask all three of you: I heard this from munib for the first time a couple weeks ago was aginated and everybody is now talking. I cover, Cloud native and I have first time heard. Okay aginative tell me about Edge native. What are companies doing to go to Edge native, and what are you know, different types of examples of what do you think they're going to get as business value of now Putting, Edge native as one of their priorities, so I'll just start with money, then go this way.

D

Sure um thank you for attributing that name to it. um The reason I called it out and I think it's everybody's familiar with what it is. First, the application attributes are different like what you have to design for, like we're all, maybe for 30 years, very good at streamlining it and we had a it workload. An IT, workflow well defined in our mind. You know from virtual machines to kubernetes to Containers, like they were like, oh with virtual machines. Does it create blueprints?

D

Oh I know how to put together an e-commerce app there's a web tier there's a database here, there's a app tier. It's a blueprint, I push it out when it comes to kubernetes I have a set of microservices I can orchestrate them as an industry. We've streamlined these I.T workload and ID workflows, we're all working to make them more agile and efficient and learning what we're discovering at the edge are OT workloads and OT workflows. You found each one of us talk about. You know, meat processing, we talk about.

D

You know transport units, and you know Road units and defense. The workflows are very different. Therefore. The workloads are also very different. The workflows are, you know, is there more? You know meat on the bone. um For me, it's like hey. How do you do traffic manager so the workflows we need to adapt to?

D

So if you actually peel this back and look at the application model, your compute network storage requirements, um you know power efficiency, you don't care about optimizing your application for power efficiency in a data center or Cloud, because you assume there is unlimited power there, whereas at the edge you can't take that assumption. You're an oil rig you're in so your power consumption needs to be super low, so the type of attributes you have to take into consideration to write an application, completely change. As you try to write this for the edge.

D

That's the reason to call out that: hey, don't just try to take a you know a data center or a cloud like technology and apply it you'll have to refactor it re. You know retool it to consider these new attributes and then to find a better way. I call it Edge native, because I have a list of 14 attributes.

D

If you guys want to follow up which, by the way, I built that you know VMware, because I've been looking at this for two years and now at Intel, but um those 14 attributes uh come from me engaging with 100 plus, you know companies globally and trying to write these Edge native applications and discovering that what we've done for 20 30 years is not going to cut it. uh Because the parameters that we have to assume and write to are completely different.

A

Got it Maryland about Edge native I know you have talked about it yesterday in your education.

B

Yeah I had my session with Ed in the audience here, and I'll actually can touch upon a little bit about the data aspects um about Edge native, so many of course covered um well. Here are the all the attributes that of course um he built for for us, but um so I think it's one of the interesting stories is that everyone's telling their stories when I look at the marketing across all of our different companies here at the edge we all want to.

B

We all want to tell the same story that you can have your applications anywhere, um and you know anytime anywhere right.

B

You can play somewhere if you'd like, but the the logistics of it I think is what we're really talking about is having a consistent operational model and management model across you know your data center, your public Cloud, your Edge, but in terms of actually deploying the applications you want to, of course build them in similar ways, but you have to take into account the different attributes of what it means to be Edge native and then of course, so um something I.

B

Just let me know I know we're talking about Edge native, but how we handle data at the edge just needs to be handled differently. Like Stu said some of it, you want to just kill the data at the edge, but in the cases of these AIML models like once, you process the inferencing, you might get additional new bits of data that the model maybe didn't process. So we have to find efficient ways to do sending that training.

B

Those training bits back to wherever the central model is, which typically will still be in the data center or the public Cloud, but then also to do the real-time decision making you need a stream or messaging queue to put it on, to be able to then process and give access to to the other applications that are running at the edge to really tell the complete story.

B

So you have to take into account not only the um you know the refactoring, but how you handle the data at the edge to like we said the networking at the edge is just it's a whole new model of how this all kind of comes together and I'm. Really excited that you know over the past year or two that edge native now really is like a proper term of how to describe applications at the edge. Let's.

A

Do you want to add anything yeah.

C

So I I actually not heard the term Edge native uh before, but um you know when I think about you know the general Trend. What have we been solving for the last couple of decades? We're really talking about distributed systems, so um The Edge has certain challenges. I talk, you know: what's your network connectivity Moody, you know power and everything like that.

C

From a software standpoint, though, you know there were certain things that kubernetes made decisions when we built things that if you talk about the edge it's like well, kubernetes is a little big and it's built for a lot of environments. So you know containers phenomenal at the edge but like all of kubernetes. Well, how much do I need- and um you know you look at the hundreds of additional services and projects that are around kubernetes at this show. You know how much of that will play at the edge.

C

So it's you know, what's the same, what's different optimizing for it, so there's a lot of hard engineering work to to adjust things, um and so you know we have multiple products to fit. You know just from you know, if I'm doing containers with Linux, you know, we've got some Solutions there, we've got kubernetes Solutions and we've got some newer. Solutions we've been talking about this week, great.

D

If you haven't heard it just blame for blame me for making it out so.

A

That's money, one question you know: in the example you gave, you know how much of time did the drivers spend or save you know in this in this example that you put together here.

D

Yeah, no, it's a it's a great question. You know I, think um we'd, love to you, know we'd love to save. You know time, for you know all of us to work life balance but I think where, as I said to you, we're applying for emergency services, so um we would like to consider that you know ambulatory Services, we've reported um saving the five to ten minutes saves lives, so the 10 ambulances in UK have already saved lives because they've been able to um in the police case.

D

The worst case scenario presented to us was a high-speed. uh You know car chase where they're all trying to you know coordinate among themselves. But then, if you have ai with these units, actually looking instead of you know following efficiently a bad uh driver, you could actually you know AI will you know? Do the traffic management block it and actually give them a GPS signal to get get ahead of them? So, but it's coordinating all the services, so saving lives. You know incidents and accidents, and you know so I would say.

D

The measure of success is really you know. I know we get excited about. You know congestion that we all live through, but imagine the Emergency Services trying to get through that congestion. So um you know all the places we're testing has been made human impact, okay, and we will talk a little bit more about this. You know next year, I think we're going to unveil a bigger thing at you know: Mobile World, Congress and we'll have our CEO Pat talk about all these um tech for good projects, which are saving lives around the world, got.

A

It okay I'm, going to ask one last question and then open it to the um to the audience to see how many questions we have, but one of the things that comes up with uh Edge is the integration of it and OT, and if I were a CEO or a CSO, or you know a lot of those folks, they're going to say, hey I'm, connecting these two, how many people are going to be able to hack into break into my systems now which are crucially running my you know, utilities or manufacturing or whatever I would like each of you to just bring up some of the things.

A

You know that you would from a business perspective. Talk to these. You know this these stakeholders and say why would you do Edge when it is going to bring up? You know it and OT together? Who wants to go first, okay,.

C

So I'll go yeah, I mean back when I was an analyst we used to say you know Edge, you know exponentially increased your surface area for attack, so that's challenging um and it's interesting a lot of times we talk about like oh, it's an edge device or a few things, one of the things we're looking at, especially in this community. What about at scale so we've talked about. You know auto manufacturers and their Fleet.

C

We had an announcement this week we took um a project that called microshift and we productize it called Red Hat device Edge. We have announced it with Lockheed who talks about their drones and if you have you know a drone that needs to fly for many hours and adjust and use the AIML technique. While it's doing it obviously they're concerned about you know, vulnerabilities and what they're doing- and you know they might have thousands of these and how do I manage that Fleet.

C

So it is something that ties into the broader discussion of containers and kubernetes, and what we're doing- um and you know we've been- we've been looking at solving that you know in the cloud in the data center and now at the edge it's part of the overall solution, so um yeah I mean the the one of the I think it and OT one of the biggest things is like the old LT world. You know you install some manufacturing device and, like nobody, will touch it ever again.

C

Well, look I mean in the IT world a lot of times we install stuff, and we don't want to update it in the software world. We all know the best way to be. The most secure is to be on the latest version of things, because you'll have all your patches and you need to do it so we're we're getting through some of those OT challenges, but it is a big Challenge and a lot of that is, you know, I'm a networking guy.

C

So it's you know, the layer, eight and nine the people in the politics um is is where a lot of the challenges is and the technology is helping enabling it. Our companies are all trying to help companies. You know move forward uh and adopt those things got.

A

It thanks thanks. You want to go Maryland yeah.

D

B

um What was I gonna say and I love. This question I feel I answered this question like we earlier. We were just talking about as we're. You know preparing for announcements in the shows and talking analysts.

B

This is definitely the question we always get, and so, besides the people, politics of the itot convergence from an actual technology perspective, like the videos we'll think about VMware, is that, as as we look at what products to optimize for the edge you know, we do have a lot of good networking and security products that can help us in particular, use cases to to teach our customers a variety of ways that they can secure um and segregate out the OT workloads if they need to, if that makes them feel more comfortable, um while still providing the the all the advantages of being able to virtualize to make it easy to deploy to do all those bits for the OT workloads.

B

um I. Think as we, um the overall advantages that you get with being able to do, the virtualization and move to the modern apps outweighs perhaps some of the concerns of how they would do it. But then also the nice thing is with the having the flexibility with the edge and the different ways that you can deploy it. You know you don't have to just have one Edge in manufacturing.

B

You could have multiple edges of different sizes, all working together and therefore- or you can also then business logically separate things out if you'd like to or if you want to add in redundancy I know at the edge for in many cases we're saying you know we want to make it easy. So you don't um you shouldn't have to over oversize your Edge for redundancy, but in cases of critical applications we can work on making it easy to still still be redundant, but still be in a compact way.

A

D

Name yeah, um you know, I talked about your it OT. You know workloads and workflows, they're different they're for how you secure them. Absolutely right. The tax services goes huge. There is a management plan issue about how do you distribute secure security to them? None of these edges potentially have large firewalls or things that you put.

D

So you have to almost tie down the you know the root of trust down to the device itself, so you have to you, know, secure the devices and they have to do a lot of zero trust Frameworks, because it's truly zero trust, because these assets are not sitting behind firewalls, and you know um uh things like that in the data center or Cloud, um you get two two concerns I.

D

Think there's one the scale where things Stu kind of pointed out to this, because um a challenge on management and pushing security policy is the scale we're all used to and we'll say we are all like. You know it's given the it's world. If you have public clouds or data center Solutions, you know, given my 10 years at VMware, I could probably count in one hand how many customers had more than 100 data centers, not a lot.

D

And then, if you look at all our you know, VMware partners- and you know- everybody's partnered- with the hyperscalers- how many regions or zones do they have on average 150 200 Max, so that the technology that has been built to scale for you know a few hundred locations but hundreds of thousands of workloads. What you're dealing with here is hundreds of thousands of locations and very small amount of workloads. So again that design framework needs to change on application distribution.

D

Once you distribute the application how's the security, you know, posture you're, going to distribute to it, and the security posture needs to be validated through a router trust which is almost built into the device, because the device doesn't have a huge firewall in front of it. So again how you write the application? How do you secure? It goes very quickly from the app to a hardware, that's deployed in the router trust buried in the silica.

D

Unfortunately, because there's not a lot anything no latitude in the middle for you to apply any types of security policy right, so um it's an interesting. So, yes, the tax service has gone High. How do you propagate a consistent security posture to tens of thousands of you know devices, and then how do you ground it in a router trust?

D

Which is you know you don't have too much to work with, becomes an interest, interesting challenge, but again, as I said, Asian native applications, the refactoring is to identify your router trust, not somewhere in the cloud, because again just two, you may have a lot of times disconnected operations right. So when it's disconnected, you can't trust. uh You know a certificate Authority in the cloud it has to be. You know, route to trust, some things on the device. So it's again your design principles, change, correct.

A

I have a question on disconnected which I'll leave to the end. If we don't have enough questions from the audience, but I would like to open it up to the audience about. You know you have any questions yeah, please go ahead, yeah right here! Yes,.

A

A

B

Okay, so let me.

A

Just answer the OT question: first, it's operational technology that is embedded into shop, floor devices or or in a utility company that is doing water flow. So that's the question was you know what is OT? Those are things that you know are physical in nature, which will open valves, close vows start a press, stop a press in a shop floor and when we put Edge Solutions, we are bringing these two together. I will let Stu repeat your question and answer that question sure.

C

I I think if I summarize your question: it's okay, kubernetes or how do I? How do I choose? How do I think about the architecture at the edge and it's actually been something we've been debating at this conference for I? Think at least four years so there's you know: k3s and k0, Os and micro, Cates, and everything like that. So red hat participated in a number of those. We created a project called microshift, because a lot of it is, you know, know thinking back to what munib talked about. There's differences in architecture.

C

If I'm thinking about you know deploying in a traditional Enterprise data center or I'm thinking about deploying in a hyperscaler The Edge has very different needs and some of the things that we think we need by default, I don't need and then there's other things that I do need that I I didn't need in that environment. So we spent a bunch of time. We've worked in the community on that project and- and that's we've recently productized that piece.

C

But it's you know we have a full spectrum of if I just want to start with containers, because I mean look, kubernetes is awesome, but you know if I just need a container I, don't need an orchestrator for that. So you know when do I go from just doing some. You know handful of containers to when do I do something that that's closer to what we're doing in kubernetes and when am I doing kubernetes.

C

So we we spent a lot of time bringing kubernetes as small as possible and try to make it as operationally the same so that you could have a very much a consistent experience between the kubernetes environment, some of your other kubernetes environment and the nice thing underneath we have. You know a common Linux platform, obviously, and.

A

So there are other two panelists yeah.

D

You know um just to add to it right, so there's no, and it almost reminds me: I'm uh I'll show my age now, I'm kind of Linux kernel contributor back in the 90s, and you know when we wrote the Linux kernel was a big impact. Then you had to do our real-time OS right, so we had to carve out a whole heap and throw it away, but you can go out there and see there's real time, always of different flavors. So what I'm getting at is based on your use case.

D

What goes into the components will vary back to your first question about operational technology. The workflows are not consistent in a meat factory to a defense to a you know. So the workflows are very different in these, because you're dealing with real life situations, which are analog, signals you're trying to convert into digital signals right so um what's more important, for you is to understand the modularity and be able to use the right set of modules to get the outcome at the edge.

D

So I think all of us are focused on the modularity of what that real-time OS and it needs to be real time too, by the way right here, if you're in a traffic signal- and you don't cut off the the green to Red in that fraction of time, then there's going to be accidents right. So uh the real-time aspect was super critical. So what you pick and choose, as the ingredients of that will be based on what type of outcome and workflowing it in itself for.

B

I was going to come what um Stu, what you said, the the fact that I mean the beauty of kubernetes is that you can mix and match all the different packages for the different areas, and that that is the massive advantage of having kubernetes. So that's something that we we don't want to be as we build for the edge too prescriptive, but we do have to start somewhere because then, if you think of it's something that keeps me up at night, worrying, in fact, I was talking about it earlier this week. Where, like?

B

How do we pick? How do we decide um the beauty of it? Everything that we're doing at VMware for our Edge optimizations across the different packages were, of course, contributing back we're listening to our customers to see. What's going to be, you know the most useful and I think we're kind of just basing our decisions and but allowing to be as flexible as possible, and then the customer can make those design choices based on the use cases, because maybe they'll use a lot of stuff.

B

That's at Job optimize and maybe one package that won't be and we'll just have to account for the extra space or computer or the different needs.

A

Okay, so I do see a big sign saying stop over there, so, unfortunately you can ask the panelists, after the end of the session, if you don't mind, but one last thing, I just wanted to say when you talked about disconnected environments, I suggest all of you look at private 5G and see how that can help Edge on the you know, on a disconnected environment with that, please give a hand to our panelists. Thank you. Thank you for having me.