Kubernetes IoT Edge Working Group, 8 Feb 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes WG IoT Edge 20230208

Description

February 8, 2023 meeting of the CNCF IoT Edge Working Group. General discussion of group activities and a presentation on AKS Edge Essentials Kubernetes for edge. The AKS Edge Essential presentation included a demonstration of device integration using Akri and an OPC UA gateway, and a demonstration of Web Assembly support.

A

uh Hi welcome to the February 8th 2023 meeting of the cncf iot edge working group on the agenda. Today, we've got three items and if we have time, members are welcome to add more uh first one is just a quick reminder that kubernetes on the edge day, is slated for April 19th in Amsterdam, but they are still open for speaking uh proposals. That deadline is Sunday at midnight, so you've got four days to get a submit in for talks.

A

So if you're interested in speaking at a kubecon event, uh go for it in the past, those have ranged from I think in Los Angeles we had 600 people register for that event in Detroit I think it may have been two to three hundred ish, so you'll get a pretty good audience and uh you might also want to consider attending because uh they're you know this is a fast moving area and there were a lot of good sessions.

A

um Next on the agenda, um we've got a notice that the edge native principles um white paper was uh got a press reference. um I, don't know. If there's anybody on the call who made it, who wants to add a little color to that, if you, if so jump in right now,.

B

um I can add a bit to that, so Brandon Wick has kind of been in charge of getting our uh first edition of our Edge native application principles out into the world um and has been working in conjunction with the cncf media team around that, and so the first wave of this is that they published it in Cube weekly, which is kind of a Weekly Newsletter. That cncf stands uh stands up every week um and then the marketing community of the cmcf is going to create a PDF um that they'll put in a blog post.

A

Okay, thanks Kate, then third up we've got um I I, hope I've got the name right. Eugene Kim of Microsoft will be speaking about AKs Edge Essentials. So I made you a co-host. If you want to share a screen with the group, you should be able to do that now.

C

A

And take as long as you want there's one other item on the agenda, but if we don't get to it, we can just uh I'll shoot out what it's about in chat.

C

Sure sounds good. Thank you so much um so hi everyone. My name is Eugene I'm, a PM at Microsoft, working on both AKs Edge, Essentials and Agri, which is a CNC of sandbox project and I can talk a little more about that later um and then I'm also here with Sean. Who is our developer.

D

C

Edge Essentials um Sean: do you want to give a quick intro.

E

Hello, everyone: this is Chong I'm, a developer, founder, aksh, essential things, yeah.

C

E

C

We also have um a couple more team members here to help answer any questions that you guys might have at the end so um yeah just you can stick them in the chat or raise your hand whatever is easier. So let's get started.

C

um So you know, as you may know, um the cloud is where rapid Innovation is happening right and on the edge. We have Legacy solutions that generate a lot of data which is chatty, and it's really about sending Telemetry to process it in the cloud, but that can get pretty expensive when there's a lot of data involved.

C

So the value of the edge is obviously to do the intelligent process in filtering at the edge based on your scenarios, so that you can control what actually makes it up to the cloud for cloud processing and, additionally, you know Edge compute creates more flexibility for you, so you can also reduce latency. So, according to some data from ground view, research, we see that we can expect 125 billion connected devices by 2030 11.2 billion Edge Computing Market size in 2022 and 155.9 billion revenue forecast for Edge Computing implementation in 2030.

C

And based on our customer Discovery and research, modern solutions are now spanning both Cloud to the edge, and this enables more developer efficiency because we can have apps run everywhere and customers would like to you know, reuse, existing windows or Linux code to get this running as well as leveraging existing Hardware since buying new hardware can be pretty costly, and so our value proposition is that we can take care of building and managing the infrastructure for users so that they can just focus on the business, value and logic.

C

C

So running kubernetes in the real world is pretty complex and costly. So first you need to stitch together multiple open source projects, tools, kubernetes distributions into a gap-free solution.

C

And keeping it your container infrastructure secure also becomes quite difficult and complex, as well as keeping all your things updated at scale, and it can be pretty costly to do all this.

C

So why do we want kubernetes on the light edge? So here we see a diagram of an industrial scenario on the bottom corner. We have a compute unit and some Factory equipment domain controllers Etc.

C

Now, if we take these up a level and Abstract it, we essentially have an expandable, Hardware compute cluster, and this acts as the central brain of the factory floor that can talk to the various domain controllers. An idea is to run kubernetes on the hardware cluster, and this allows you to scale out your central compute unit and talk to the domain controllers through the open source project Opry, which acts as a connector between your Brownfield devices or sensors to your cluster.

C

And at the top you can run your containerized workloads as kubernetes pods on your cluster, and now you have kind of like a horizontal scaling orchestration and you can also scale upward Down based on your needs.

C

So what's AKs Edge Essentials, so it's a lightweight platform solution for running containers, applications at scale on light edge devices, and we saw the challenges of commercially deploying kubernetes at the Edge by providing a Microsoft supported software stack from kernel to the cloud and windows to Azure covering your solution from top to bottom.

C

We also deliver cncf, conforming k8s and k3's distributions, and these are managed and automatically updated for you to ensure that your kubernetes stack is the latest ingredients and we also support both Linux and windows. Containers which can be deployed on both single or multi-machine clusters, and your deployments across all clusters on the edge more Cloud can be easily managed with Azure Park, so that you know you can focus your energy on the applications and features that you want to deliver to your customers. Not the rigorous work that it takes to manage a kubernetes infrastructure.

C

So with aksh Essentials, it's pretty easy for you to get started with your on-prem kubernetes deployment. So you get a Microsoft managed lightweight kubernetes distribution, which can run on something as little as four gigs of memory and two virtual CPU compute and we've also simplified the installation experience via an included Microsoft managed VM for easy installation and configuration.

C

There will also be ongoing monthly security updates from Microsoft to keep your Linux and windows VMS um secured over time.

C

We also provide automatic updates to stay aligned with your Cloud kubernetes version and you can choose to run Linux windows or both containers when creating a kubernetes clusters on AKs educations.

C

So now, let's just look at the overall technology stack of how it works. So at the bottom you have your light edge devices and then, on top of that you know you would run a Windows client host of some sort. um We lean on Windows Enterprise and then you can install AKs, Edge and Essentials, just like any other application through an easy MSI and so um inside.

C

You have your Powershell commandlets and agents that enable the provisioning of your infrastructure and then your cncf performance, kubernetes platform, and then the user can just manage your cluster using either standard Cube, CTL or Azure Park. And on top of that you know you would deploy your Linux and or Windows containerized workloads.

C

um This is a bit more of a deeper dive into the architecture. So again we have your hardware and operating system and AKs Edge Essentials consists of the Mariner VM and you know optionally, the windows VM.

C

If you want to run Windows for glues, but on the Mariner side you would have your k8s or k3s control plane and then any other Azure containers like flux agents or the art connected cluster agents would also sit on the Mariner side and you can actually have Windows native apps running on the host that can actually talk to um the VMS running inside AKs, educational. So, for example, you might have a retail setting with a UI.

C

That's a Windows native app that reads: Telemetry from machine learning, workloads running in the Linux side of your cluster on the same device, and this interop piece is what we'd like to see users adopt.

C

And so now, um Sean's going to talk a little bit about the k3's windows, enable Network that we did um Michelle.

E

uh Sure uh Eugene can you yeah? Thank you. uh This is Sean and, as we engineer, AKs essential one of our achievements to bring up k3s for Windows support. As some may know, uh KCs is a kubernet implementation which is optimized for memory and disk footprint. It comes with many interesting architectural decisions to distinguish itself from the Upstream KAS, for example, instead of virtualizing the kubernetes core components as running Parts, k3s fold them into one single process and they run them in multi-thread environment.

E

This decision May expose the disk footprint and the memory usage more compact when it comes to packaging all of the required runtime binary package in a single binary, and it is executing all of them at runtime when needed. It is also natively supporting to manage precache user container image. It is very useful for the offline deployment.

E

Less but not least, k3s is engineered in a way to export different functionality by calling k3s space at XYZ, which is they call a multico binary and, as our teams start supporting, uh we would like to preserve all of the good feature of k3s as well, and this is this slide shows an illustration how we put things together for Windows port, and this effort was not start from the scratch.

E

K3 has already made a few a skeleton change to make it ready for someone to continue to grow across platform porting to highlight a few things we made. For example, we update and pick up the correct set of cni binary and the content Edition for Windows platform. We also update the default flannel setup vendor to support Windows cni initialization, and we fix a few uh file paths convention issues and update the view scripts to be more like Windows compatible using.

E

Can you move to the next slide, and we hope least ever is not a work only limited to aksh essential. Therefore, we commit to the community to bring back and Upstream those change back to the KCs project and to make it more sustainable support for long term. We also commit to work with the project owner to improve the windows validation.

E

Please stay tuned for our coming activity for k3s, and thank you all. uh Let me hang over the control back to Eugene.

C

It's going to be simple demos using AKs, Edge Essentials, and this first one we're just going to demonstrate like a really simple um laws and workload that spits out hello, hello worlds.

C

um So here I have a config Json, where I can configure the cluster that I want. So I want a Linux node with four CPUs um four gigs of RAM and 10 gigs of data. So now I will pass in the config.json as a parameter and do new AKs, Edge deployment and so I'm just going to skip over a little bit. It takes I'd say like two minutes for the cluster to be deployed.

C

So now our cluster is deployed and I can do compute control. I can pause, see all the necessary pause are running, and now we actually have some samples that are public, um so this Powershell script will enable you to run walls and workloads on AKs Edge. So here I'm just running that Powershell script.

C

It'll download, you know the shin version that you want and do all the configuring of the files and environment variables for you just.

C

And I can drop a link to this sample the chat after so now. That's done I'm going to apply the runtime class um for both um Spin and slate and then I'm also going to deploy some simple workloads to see if it's working properly.

C

So now we're just going to wait for the pause to spin up and you can see that the spin and slide workloads are up and running now, I'm just going to get the service of the port um so that I can curl it and see that it's working correctly. So you can do invoke AKs Edge command to just do your Linux commands on the Note, um so here I'm just copying the service IP and Port hello.

C

For slight and there you go, it spits out hello, and you can also do the same for the spin. So just replacing the IP with the spin IP.

C

It says hello world from spin and then I also have one more quick demo of using AKs Edge Essentials with Aubry, um so here I'll be demoing, discovering OPC UA servers.

C

So I already have you know a k3s cluster up and running here and now I'm just going to update my help, charts and use home to install all green with the correct OPC UA configurations. So here I actually already have um some container instances running that spit out simulated, OPC, PLC values and then I'll actually put in the fqdns into the URLs so that they can discover the correct server. It's it's foreign.

C

Pods are running and if we get just give it some time, it will discover the servers and create an awkward instance resource for each of them. So, since I have two servers running, we see two augery instances here and now: I'm just going to apply an anomaly detection app, which will use a broker to take in the values from the servers and highlight any outlier values.

C

So I just take the port of the application and the node address.

C

And I'll just put that into my browser.

C

That'll take a couple seconds.

C

And now you see, the values are being logged and any outlier values will be highlighted in red and I. I can also link to the sample of this as well in the chat.

C

Okay and that kind of wraps up my presentation I'd be happy to open it up to any questions that anyone may have. Thank you.

A

I guess I'll start is: is there a link where somebody could get access to try this out? um You know, I assume that tried out would mean get something that you could run on your own print Hardware to uh emulate actually doing this at Edge, yeah, on-prem, environment.

C

Yes, we actually have documentation which I will link here, and you know the demos that I showed I just ran it on my work laptop. So you know it's as easy as that to just like: try it out on your phone um actually.

A

And in terms of supported uh platforms, um I'm wondering how small this would go. You know what, if you were to put it on an edge device. Is it strictly x86 at this point or do arm things like Pi's work uh and what would be the memory and number? Of course one would expect to run a minimal kubernetes cluster at an edge location.

C

um Francisco, do you want to maybe help answer the question.

F

Yeah I can help with that. So, first uh regarding architecture. So right now we only support 86 right, but we do plan to support our arm.

F

64., uh that's gonna, probably come late this year uh and in terms of the um the horror requirements um right now we we at least uh you know a device that had at least four GB of physical memory right and probably at least 2.5 GB of free memory and then two virtual CPUs, with a clock speed at these 1.8 gigahertz we're working to bring that a bit you know down, but that that's right right now and then and then just one more thing there this this space, we need at least 14 GB, free.

A

Okay, that's a storage right.

F

A

And I'm wondering how broad this product offering is? Is it strictly kubernetes or are there aspects of network connectivity and say a container image registry uh involved with this as a a packaged uh unit for users to consume it.

F

So do you mean like what's included right with the product like? If it's just you know the kubernetes distribution? If there's anything else,.

A

F

Well, so so right now right issues it so that we you can, uh you can add right so right, as you know, kubernetes right, both k3s or KAS, and what we already have there also integrated is Arc right and you're familiar with Arc right, but it's this kind of service where you can connect right and manage your devices and your kubernetes cluster directly from Azure and that's already kind of integrated as part of this product.

F

So if you want it's just an optional thing right, so if you want to opt in right, you can go ahead and do that and you know manage everything through the cloud. But if not, you can just do all kind of on the edge right using Powershell or using whatever other kind of Windows tool is the the available.

A

Yeah I'm just imagining a scenario where out at Edge, you sometimes have intermittent connectivity, but you'd still expect to be able to run even with loss of your Uplink and in uh many things that people put together for Edge operation. They would include things like running a local oci registry so that, when apps start to restart, they would pull their images from a local Source rather than requiring Cloud connectivity and I'm just wondering if that sort of thing is part of the package or something that you would you'd add on. If you needed that facility.

F

Yeah, that's an interesting question and yeah. We're kind of aware of you know if these kind of offline scenarios right now we do support offline deployments. So basically you could do that right and you could have your private oci right. Just if you have a generation in your private Network and just you know, download everything directly on your own Prem right uh when they're providing those sort of products.

F

Although because we're kind of political, like you know, cncf conforming right, kubernetes right, you can just go ahead and do that right as long as you're using you know, oci content registry that should work.

F

um The other thing is we we're working to bring other features that will kind of enable this kind of scenarios like, for example, multiple nics right, so that you can have you know, for example, an online Nick and then having a an offline Nick, for example, for a DMC kind of scenarios right um but yeah we do support offline offline, Employments and, and the other thing is like regarding other features like, for example, updates right or management right, because this is this is Windows right. uh You can always go ahead and use whatever Windows tooling.

F

There is right for offline kind of uh Windows requirements right so, for example, if it's, if you want to apply the updates right and you want to do offline updates right- you can use this all the you know, AKs updates are done through Microsoft fabric right. So, if you, you know, if you're, using whatever wsus or other windows mechanisms, are for applying offline Windows updates, then you can use those same update mechanisms right for these kind of a clusters.

A

Okay thanks, you answered my questions.

A

Anybody else who has questions please jump in.

G

H

Have no question: oh yeah, go ahead, all right, yeah sure thank you. um Hi uh I actually really enjoy your uh demo. uh Actually I I used to work at AKs, so for a few years under brilliant So like um yeah, he feels really really like uh like home. You know uh actually uh I have a question.

H

So uh what's the recommended, uh inter cluster networking, so I think like between the Clusters, like, usually, you have a lot of like the you have to set up on some VPN or whatever yeah that's and uh you have to make the like the subnet right yeah. That's also always a pain point, and the second question is like the what's: the memory uh requirement for the windows uh k3s nodes.

H

F

uh Okay, so first, let's let's go ahead and turn on networking, uh so we do have two deployment types right. The first one is what we call the single machine cluster right. So this is just you know. You have one machine that you can run either. You know you need to run a Linux zone and you can add a Windows now Place optional right when we do that right. What we end up doing is we are using an internal virtual switch and we we set up.

F

You know we created internal virtual switch for you and we set up the IPS right and then that right, so that you can, you can you know the external network right, but all these miners with my internal virtual right.

F

uh The second option is what we call the scalable cluster right so as well. You have multiple devices right running, maybe you know one or two knowledge each of these devices and they need to communicate between the different nodes in different devices right so that you know there. What we use is we use an external virtual switch right, so each of these devices right will have we'll use an external virtual switch and look.

F

You know you know we'll connect the different VMS, the Linux VM on the Windows VM to that external virtual solution and communicate directly to the external network um when it comes to more specific kind of you know. uh Networking parameters like like we support static Mac address. We support mtus, we support proxy uh right um and I. Think the other thing is custom DNS also, so we do support all of kind of networking. Vpn is not something that we do support right now, but me.

B

A

Have your VPN.

F

Right correctly, a regular external network so.

A

For this external virtual switch, what's the technology there are you counting on it running uh with some kind of Open, Source, virtual switch or in a hypervisor environment? That would provide that or some other thing.

F

Windows right and we're doing all the hyper we're using hyper-v hypervisor right. So this is a hyper re. Virtual switch.

A

F

um And then the other question was regarding the memory requirements right um all shared here. You know they. You know the system requirements the page on the chat right, but basically, as I was saying, is you need at least uh a Windows machine that has 4GB of ram right and at least 2.5 GB free?

F

If you want to know you know if you want to have an R connected cluster and you want to get UPS right, you know with with Arc, and you know with flux, then there's some more memory requirements and you know how you will need at least an 8GB uh machine right with at least 4.5 uh free rumor.

H

Got it? Thank you.

G

Yeah, my question is related to the network. So can we still buy into the cni to to the work? World am I right, any cni as long as implementation supports container network interface.

F

Vincent, are you there I'll refer to you or Sean Maybe.

G

G

You guys use uh you mentioned k3s right as a back-end, so.

E

Yeah, so for uh it is essential where we provide two uh distribution to choose one uh for the KAS, uh the default one. We recommend to use the Calico and then for k3s. um We use it uh before implementation so which is the Plano um um yeah. So that's the the cni indentation count with the AKs essential like by default.

E

um Well, we we don't officially like recommend to uh pull it and use it to uh swap the cni implementation. And then maybe it's not that easy to that, because um the the VN, the note itself is, the composition is managed by us yeah. So back to your question, I would recommend you just go with the default implementation, but for experimental purpose. Sure.

G

E

G

E

um Yes, for KS, which is calico yeah, okay,.

F

And that's part of if you go some requirements Matrix there and it will explain, you know, which is you know, which is experimental, which is probably for each of the different distributions.

A

Okay, so another thought related to the cni. Plugins are the storage plugins, you know suppose I wanted to run a stateful app that would use um CSI storage are any of the plugins supportable or are there some bundled? You know in particular um I'm thinking. If I wanted to use an NFS, NFS mounted storage, there are some kubernetes distros that run on a with some of the minimal kubernetes. Don't include the kernel features that would enable that out of the box, so I'm wondering what the situation is. There.

F

Yeah, so we do right now we support right uh for the Linux node. We support the local path, provisioner right uh and we support NFS. Yes, we do have the you know the kernel and the Linux packages needed to do that.

F

Okay, um when it comes to other kind of, let's call it external Provisions, like Longhorn, you know open EVS group like we still haven't tested those and I think we have some some packages missing, but but local path and NFS work for Linux and we're working uh support that to for win for the windows notes. Also.

A

uh What about the Linux distro is it? Is there one bundled with use of this, or is it open to the user to provide a Linux of choice to host these nodes.

F

No, so it's always it's always. You know that Linux VM is always based on our own Linux distribution inside Microsoft uh and the VM is it's a big kind of particular because it has a read-only root file system and has no package manager. So, basically you cannot add packages. Issues like you know. If that VM is only used, you know to run. You know these containers right uh and kind of you know the way kind of one of one of the reasons for your advice. We need to provide.

F

You know a good update story right um and you know we provide. uh You know from kernel right from from Windows. All the way to I guess actually provide that support and that update mechanism.

B

So, by providing.

F

That you know kind of no package manager. We only refill system, we can apply. You know this kind of kind of a b updates right uh uh without issues.

A

How does a Linux update work in this scenario? Is it managed, externally, by the um the uh AKs control plane, that is cloud hosted, or uh are there extensions to kubernetes? That would do this on a local basis.

F

Yeah, maybe Sean Vincent. Can you go I'll be into the details? There.

D

Sean, do you want to take this? One I can I can add uh later on I'm.

E

Sorry, uh so is that a good question like how we're doing update.

A

Yep yeah so like, like suppose, I'm a user and I want to either initiate an update or monitor the updates. Where do I go to do that is that at some Cloud control, plane and yeah, is it automatic, or is this manually driven just curious about how that this, these OS at Edge node updates, would work same with kubernetes updates at the edge node.

E

Sure yeah so, uh as usual mentioned like in early slide like um the AKA essential, is packaged in the windows MSI installer, so it will be like installed as a regular application and then in this case um the windows will regularly check with the Microsoft update and if there any update, then we will just like download a new uh update package which is another MSI and to deploy to your machine. But at that moment uh the your cluster still remain the original state.

E

We will not touch your um cluster like out of your consent and the end user. When end user is comfortable and ready to update the cluster, uh they can run the partial commonly and to confirm there. Is there any update and if it is then trigger update and then the process can be automated if there is some other it infrastructure, but in general, that's the the process, how we, um how we trigger the opting yeah.

A

Okay, so let me just let me repeat what I think I heard just so I'm sure I understand, so you do something like uh Microsoft hosts out at the edge Windows hosts. You would probably use something like Active, Directory, Group Policy, to configure policies of how the updates and would be policed and monitored and downloaded, but they get staged downloaded and it's a manual operation that one could use a Powershell script or API to trigger the actual update. So it do I understand that correctly. Then,.

E

Yeah, that's correct yeah you can combine with other like um group policy or under uh Mage I mean Windows host the management solution which supporting the the the the MSI management, um yeah and.

A

If I did one of these updates and found I had a bad day, is there a way to roll these back then through the process.

E

um So for now, um if the update is um going bad somehow doing when you like trigger the update and we were doing a rollback, but if it's already finished- and you want to go back to n minus one and yeah- that's not something. We breaking right now, yeah well,.

A

It if it works with the standard, wouldn't those update, it's a pretty large Hammer, but in theory maybe you could do a system recovery, but that would recover the entire host, not just this portion, so that if you were running other things there uh like I say it's a pretty crude tool, but it might work okay. Well, you answered my question thanks.

A

D

G

I have another question: uh it's not a question, but I would like to have comments. So you mentioned that you guys use as a back end k3s right.

E

Hello, yes, yeah. We yeah.

G

What was there any like other like consideration for, like add a framework compared to k3s.

E

um Yeah yeah, when we were um doing the early research, were some other um options like on the table but um yeah at the end uh because of the maturity of the case, us that become one of the yeah, the one uh we finally like making a decision to go East uh for the yeah for the the memory and this footprint, optimization yeah.

E

Does that answer to your question.

G

That's the major requirements right of memory, optimization and yeah.

E

Yeah yeah yeah yeah.

G

But I think I was considering I I have played with k3s like years ago and uh that works okay and, as you mentioned, like a memory footprint and CPU consumption that pretty well optimized uh I can say that. But you know like something: I consider is like architecture, so the k3s is fully compatible with kubernetes, because the problem could happen because the if the application container issues in classic config API or something realized on the kubernetes API and it breaks right, but because k3s does not support a fully compatible API.

G

So yeah I was considering a death situation and that's the architecture. So did you have any like a consideration about that? Using.

E

G

E

For k3s um for each release, they do pass the um equivalent components testing. So as long as the the the usage is under the components, testing I think they are all support yeah.

E

But maybe you are talking about something else: yeah. If they, you are learning something beyond the com. The the general complement test, then yeah. Maybe there will be some um differences between KS and k3s and that you have to uh thinking about uh yeah, which is true. We are using and you need to use the compatible whatever compelling method to to to to talk to the cluster yeah.

A

I'm kind of curious as to what aspects you think are not compliant because the the cncf definition requires you to pass a Sona boy conformance test in order to even call it a kubernetes, distro and I. Think k3s is on the list, but there are often other aspects where they leave things out that people have grown to expect. You know that are are shipped uh in the box with a lot of other distros, and maybe it's something like that. But could you expand on what aspects you're talking about as having had issues.

G

This is something something actually happened to us is like. uh We are trying to use uh kubets I, don't think that is that's the policy conformers test for the kubernetes, but the with kubets some cni doesn't work because it depends on the kubernetes API.

G

That breaks the architecture right, so I was just thinking. You know if that could happen with using this framework or something that's the background. Okay,.

A

Did you answer your questions.

G

And thank you Kate. You just posted that, like they support k3sk, 8s yeah.

A

Yeah: okay, when you say they support uh Ka k8s, is that like kubernetes, just straight out of the box from the open source, repo or something else.

B

Well, I, don't think I'm the best to answer this. But yes, when you install you can choose cadis or k3s KDS being based on yeah Upstream kubernetes. um They package themselves.

F

Yeah, that's correct, yeah. Basically, when you install you have two different MSI, so one specific for k3s and the other one for KDs right and we saw you just choose that um and.

G

F

For the k3s, it's also, you know we're using the Upstream version of it for the Linux part of the you know, Rancher, k3s, right uh and then the windows. You know, Sean explain it where packaging ourselves that and we we start planning to to Upstream. That also.

A

Maybe this is uh kind of news, so you don't have it it's a track record yet to give numbers on this, but I'm curious as to how you do tracking, with, if you're dependent on uh k3s there's a process where new releases come out of the kubernetes project itself and then there's usually some latency between when Coburn the kubernetes pure open source.

A

Upstream project comes out with a new release and when the distros come out with the release that have those same features and then, if you're, uh repackaging, k8s I might guess that perhaps there's even more latency and I wonder if you could give me a feel for you know what your tracking performance is both for uh coming out with new releases and coming out with patches when CVS are announced and fixed.

D

I can try to take that. um So, as you note, there are a few different AKs products in Microsoft, right AKs in account AKs, HDI and now AKs actually potential, and we try to align our Live support cycle for kubernetes aligned with ATS. So you uh we are going to follow the same uh life cycle as AKs in the cloud. So that's how, uh in the long term, that's what very broadly aligned.

D

As and the second question regarding cve as Francisco and Eugene mentioned, uh we are uh the the VM that we use is based on CBR, Mariner and Microsoft has internal tools to check the series and we have an SLA to adjust all the CV that is published to the database. uh We are using the same method to track the CV for key fish right. So uh when the CV is made available on this, we will we will patch it uh uh SRA and the different update for Microsoft updates foreign.

A

Anybody else got questions.

A

uh I see a request in the chat if uh you could share the slide deck or slide decks. If it was multiple speakers who were sharing.

C

um Let me um put it in subpdf and I can post it at the slide. Okay,.

A

That sounds great.

A

Okay, well, uh if there aren't any further questions, I'd like to thank you, that was a great presentation and I look forward to uh going and following some of those links and checking this out for myself, it's inspired me to have a lot of curiosity. Go on to try, try those things.

C

um For robbing us and um if you have any like questions while you're going through it, feel free to you know, slap message me or whatever: okay.

A

uh You know you did demo connectivity using acri with the OPC UA is all of those is getting into that level of detail in the documentation now, so that I could reproduce even that level of uh interaction. Yes,.

C

A

Okay and uh feel free too, to let this group know if that that, if it's getting G8 I think often people have a lot of unveiling events and things and kind of Hands-On workshops. So if you in the near-term future want to drop links to those kinds of things or YouTube videos or whatever you have into this group's chat uh stream on slack, please go for it because I think a lot of people would be interested in getting that material.

C

Yeah that sounds great I'll. Definitely keep the group updated.

A

Okay, if there are no other questions, the final item on the agenda today was I wanted to bring up this ongoing work. We've had recently, which is compiling a list of Open Source.

A

What we're calling Edge native applications so I posted in the chat, a link to this spreadsheet right now we're we will probably eventually publish this on GitHub or we've even contemplated uh talking with the leadership in the cncf about perhaps getting an edge native landscape going, but we're at early stages of just trying to phase one was just to figure out what's out there and it turns out.

A

It actually is surprising to me that if you open the link to that spreadsheet, we're now up to 150 open source projects related to Edge, and even that has been called where I made an attempt that, as I was discovering things if these open source projects are stagnant, meaning, like they've, had no contributions for a couple of years or in some cases, sponsoring foundations have indicated that these things are being retired.

A

Those kinds of things are not on that list, so it is really pretty a remarkable Testament to how rapidly growing interest is in this Edge uh space that there are so many open source projects and uh we'd like some help on that, so that, if we're missing any of these or some of the data is wrong, please just go use your credentials as a member of this group, you should you'll automatically get edit access to that spreadsheet. Help us um continue to polish that and complete it.

A

uh The next phase that I think we've already got enough critical mass, that what we wanted to do was classify these projects into categories, for example, perhaps the most crowded category there is kubernetes distributions. You know they're.

A

We already know that kubernetes itself, you know for public cloud I think has 60 some distributions, but the reality is that users kind of vote with their feet and don't want to use the unpopular ones and they kind of choose to use one that is getting more attention and in the edge space um it's a it, isn't 60 of them, but I found the number of them out.

A

There was surprising uh I would expect that as time goes on, people consolidate, apparently just like this eks Edge Essentials chose to align with an existing kubernetes, discerol I, think maybe that will happen with others in the future too, that the world does the world really need 20, different Edge, kubernetes distributions, probably not, and at some point in the future um there will be natural selection going on to where that number gets reduced, and we want to look at this list of edge native applications, divide them into categories, so that somebody who needs a tool doesn't have to go manually.

A

Read 150 entries in a spreadsheet, but maybe can say, I need an AIML solution for Edge or uh you know, I need a in an iot Gateway for Edge and look at what's out there so that you can join in the efforts already being made by others rather than Reinventing wheels that don't need to be reinvented. So anyway, that was the agenda item that I had just to call attention to that that this is still ongoing.

A

Please take a look at that and if you want to contribute your own knowledge, uh please go in there and help edit that document, uh that's it for the items on the agenda. Does anybody else in the audience have any last minute ads uh remarks, uh whatever uh just turn on your mic and speak.

B

Also, um just to let everyone know these agendas flow each time. So if you know someone that would be interested in presenting or there's a certain topic, you'd be interested in hearing about um feel free to, let us know or reach out to the person. You would like to speak. It's very Word of Mouth to get speakers to come so always open to directing the agenda where people are interested. Yeah.

A

Even if you don't know a speaker, if there's a topic, you're curious about okay, the leadership of this working group.

B

A

Willing to go do a little research to go, find who's out there on the topic. So we want to keep this interesting to users. uh That is the goal, so uh just kind of things that you want to learn about, go, throw them out here. You can either speak up right now or leave a note in our slack Channel suggesting that you would like to see a speaker on this topic, whether you know the name of that speaker or not foreign.

A

I guess we'll close this out, then the final reminder that I gave at the beginning uh speaking proposals are still open for kubernetes on the edge in Europe for the until Sunday. So if you want to talk about something, please consider uh submitting a cfp uh back conference is coming up in April. The scale conference in Los Angeles is coming up in March and has a lot of edge related sessions as well.

A

um I think that the the Linux Foundation announced some cfps for open source Summit in Vancouver that is coming up in the first half of this year as well. So thank you. Everybody for attending and we'll see you again in two weeks. Bye.

G

C