KubeCon + CloudNativeCon Europe 2024 | Alice Frosi & Fabian Deutsch | Red Hat

Clips
News
More from KubeCon + CloudNativeCon Europe 2024

Fabian Deutsch

KubeVirt Maintainer & Senior Engineering Manager, OpenShift Virtualization

Red Hat

Alice Frosi

Principal Software Engineer

Red Hat

Virtualization meets Kubernetes: Breaking down the progress with KubeVirt

Prompted by factors such as Broadcom Inc.’s acquisition of VMware Inc., organizations have found themselves on the hunt for tooling to ease their migration of virtual machine workloads into containerized environments.Born from that enterprise need is KubeVirt, which brings a unified development platform to build, modify and deploy applications residing in both application containers and virtual machines in a shared environment.“Right now KubeVirt is reaching maturity, so we are focusing more in cleaning up the code and getting more people involved,” said Alice Frosi (pictured, right), principal software engineer at Red Hat Inc

play_circle_outline Introduction to guests Alice and Fabian.

play_circle_outline Enhanced Stability and Seamless Upgrades: Unveiling KubeVirt's Updated Code Base and User Guide

play_circle_outline Focus on AI workloads and the use of GPUs in KubeVirt.

Info

Alice Frosi & Fabian Deutsch | Red Hat

Fabian Deutsch

KubeVirt Maintainer & Senior Engineering Manager, OpenShift Virtualization Red Hat

Alice Frosi

Principal Software Engineer Red Hat

search

(bright upbeat music)

>> Good morning, Cloud community, and welcome back to spectacular Paris, France. We're here at KubeCon, CloudNativeCon, CNCF's flagship European event. My name is Savannah Peterson, delighted to be joined by my fellow analyst, Rob Strechay. Rob, did you have any pastries this morning?

>> I did not, I did have a baguette though. (Savannah laughing)

>> You had a baguette this morning?

>> I had a baguette this morning, that was my breakfast, so.

>> I'm actually a little jealous, I might-

>> Yeah, it was fantastic.

>> You didn't want to bring a baguette for the rest of us?

>> I was on the walk. (Rob laughing)

>> Oh, it was on a walk. All right, all right, all right, well-

>> On the move.

>> At least I'm virtually having a baguette right now, 'cause we're talking about virtualization on this next panel. Very excited to welcome our two guests, Alice and Fabian. Thank you so much for being here, how you all doing?

>> All good, yeah, I'm glad to be here.

>> Savannah: Is this show a big event for you?

>> Yeah, sure-

>> Savannah: Yeah.

>> KubeCon is like the event. (Alice laughing)

>> Savannah: It is the event.

>> Yeah.

>> Yeah, we're also really big, super deep nerd fans. Fabian, you were on stage today.

>> Fabian: Yeah, indeed.

>> Tell us about that.

>> Tell us about-

>> Tell us about that. You're on this stage now, and we want the highlights.

>> It was, like, I feel, you know, it's great, right, to have that opportunity, and to have such a great co-speaker who can't be here right now, but like, it's amazing, right, so you-

>> Savannah: It is.

>> You have somebody, they've adopted KubeVirt, it solves a problem to them and then you can get into such a stage. I mean, it's like wonderful-

>> Rob: Well, it was Goldman Sachs who was on-

>> It was.

>> Talking about how they're using KubeVirt to actually move virtualized machines, or virtual machines, into their Kubernetes environment. Which I think is, I kind of talked about it yesterday, how I was surprised we hadn't talked about it, and I totally forgot you guys were coming on today. But now we get to talk about, but-

>> Or did you forget? Subliminally I think-

>> Maybe.

>> You were definitely teasing this interview the whole time.

>> I may may have been teasing this interview the whole time-

>> I love it.

>> I think again, it's such an important topic I think, and a project in general, because of the changes that are going on in the industry with Broadcom buying VMware, there's a lot of service providers, for instance, that are looking for other options for doing hosting of virtual machines and things of that nature because of the contract changes that have gone on underneath there. Do you see a surge of people being interested in KubeVirt since-

>> Good question.

>> All of this has started happening?

>> Sure, I mean, even here at KubeCon, we got many questions how people can transition from VMware to KubeVirt, if KubeVirt is stable enough to be used in production. So we have seen a lot of interest in the last month so.

>> And I bet you're not surprised by that.

>> I'm not, I'm not. (Savannah laughing) So what's new with KubeVirt? What's been going on?

>> So right now, KubeVirt is like reaching kind of maturity so we are focusing more in clean up the code, get more people involved. So there has been many initiative to improve the code base, to clarify more the ownership. There has been also a lot of improvement in the user guide, as many people are joining the community, this is of course the first entry point for the users.

>> As Alice just said, right, so we're working a lot, like, you know, with KubeVirt, we had a race, right? I mean, with KubeVirt, we want to provide, we are providing a virtualization solution based on Kubernetes, and just one addition to what you said about Goldman, like you know important is, you know, Kubernetes is becoming the infrastructure and VMs' become one workload that's being run there, right? And KubeVirt is enabling to do that. So we've been working with KubeVirt, right, to enable all these features that you expect from, you know, these virtualization systems that are out there today, and we've achieved much of that over the past few years. Like some big things like hotplug for CPU memory resources are still, they need to GA. Memory overcommit, higher density, that's something we're working on. But we're switching to, you know, being even more stable, right? We provide seamless upgrades, but we want to be more stable. We're working on, you know, making easy to maintain. Because if you switch KubeVirt, you can rely on it, right? And that's what we want and have to live up to, so we're changing focus a little bit, while not, I mean, while continuing to focus on innovation as well, right? New features have to go in, but the foundation needs to be solid.

>> Yeah, another interesting thing is usually virtual machine are hard, they have many options. So we have tried to simplify the deployment, so there is a new feature, and-

>> Savannah: Super important, everyone wants to decrease complexity.

>> Yeah.

>> Fabian: Yes.

>> Savannah: Yeah.

>> So for example, we got also many questions about Windows, how to deploy Windows VMs, so you can define, actually, if you want to run a Windows VM, there is like something called distance types, and this helps user to deploy a virtual machine. So this kind of template that would set all the option for you, so this, again, to make it easier for user to deploy VMs.

>> Everyone wants that easy button, we talk about it a lot-

>> Yeah.

>> And especially when it comes to Kubernetes, especially when we're talking about AI and everything else. You know, you and I were talking about your daughter's senioritis earlier-

>> Yeah.

>> And in thinking about graduation, where's the project maturity right now? Fabian, I'm going to go to you on that one.

>> Yeah, so, you know, I think we joined the CNCF incubator, in 2019, and incubation 2022. 2022, yeah. And that was always a good reflection of the project's maturity, I would say. So I think it's just, you know, the natural next step to now work on graduation, it totally aligns to what we're doing, like on a day-to-day basis. Like working stability, finishing the, as I just said before, right, so yes, graduation is definitely something we're working on this year. We're working actively like to reach out to end users, to more platform providers, like we have a lot of platform providers who are using Kubernetes and offering KubeVirt on top as a service or, you know, using it by themselves. So we're working with them to simply complete that whole graduation. It includes security reviews, it involves governance, and involves end users and other adopters.

>> How many contributors are on the project?

>> We have 150 contributors. (Alice laughing)

>> Alice: You've prepared today, so you-

>> Yeah.

>> Alice: You know the numbers. (Rob and Alice laughing)

>> And I'm even more about the companies, because the company ratio increased over time-

>> Savannah: Yeah.

>> It's always good for a project, right, to not rest on, you know, few companies. So it's great to see that more companies become interested, regularly contribute to KubeVirt, so that's just great to see.

>> Do you see people, what's I guess the ratio between people who are vendors contributing, versus non vendors that are contributing to this, do you think?

>> I would like to take that offline and work with you.

>> Rob: Okay.

>> To digest the data. (Rob laughing)

>> Rob: That sounds good.

>> I think like, you know, KubeVirt is for sure very interesting to platform providers. So vendors, right, you could say, and they're for sure, I would say they're more dominant, right-

>> Alice: Yeah.

>> In our contributors.

>> Alice: Definitely.

>> Rob: Yeah.

>> But some end users also contributing, right?

>> Yeah. And where are kind of the requirements coming from? Is it people saying, "Hey, this is what we've seen in KVM," "this is what we've seen in VMware," "these are the things that are missing." How is that working with the community?

>> So I mean, of course people coming from traditional virtualization platform, they expect certain feature to work, like for example, hotplug, as Fabian mentioned. But we are also getting requirement like from container perspective, like, I dunno, like selecting certain environmental variables or passing certain kind of files. So I would say it's kind of a mixture, but yeah, a big part of the requirement is for sure coming from virtualization, the virtualization world.

>> And it's leveraging, one of the other keynotes this morning, they were talking about the storage aspect and it was, it's leveraging CSI to be the connection back to storage. And is that, are you being able to leverage a lot of the different things that have been done for, you know, things more ephemeral container types, but for something that needs to be more persistent?

>> Yes, definitely. So we, KubeVirt doesn't make any difference. So between storage, storage provider, so if you have a CSI plugin, it can work with the file system, and it can also work with block devices. So yeah, generally if you have your CSI plugin, it can perfectly integrate with KubeVirt.

>> And can I just add something?-

>> Please.

>> I think that's true for so many aspects, right? I mean, why did we build KubeVirt and Kubernetes at all, right? Because Kubernetes is providing so much, right. From an infrastructure perspective, CSI, CNI, but also all these other tools like, you know, patterns for observability, patterns for our deck, right? Policy enforcements and such, that stuff, so that's why we built KubeVirt on top of Kubernetes, to leverage it. Why should we reinvent it again for VMs? How can we build such a wonderful ecosystem that we have with the CNCF ecosystem?

>> Yeah.

>> We could not build that for VMs only, right? So like by embedding VMs into Kubernetes, we can simply benefit from that ecosystem that's providing so much, it's like.

>> I mean, it makes sense.

>> Yes, yes.

>> Everything makes sense. So what's next? What are you going to build next? What are the contributors going to make next? What sort of features can we expect?

>> What's on your roadmap for this year-

>> Yeah, exactly, yeah, give us the preview.

>> So live update of resources is of course a big topic, so we just added a CPU and memory hotplug. So we were already supporting storage and network interfaces, so we can expect also, in the next future, to have additional feature in this direction.

>> Do you see different workloads coming to KubeVirt that are non-traditional VM workloads? Like some of the AI stuff, some of the database stuff would make sense to me, which is not a, I know you can do persistent databases on, and I sat in a Postgres persistent database talk, data on Kubernetes day on Tuesday. Do you see those types of workloads really being driven to KubeVirt as well?

>> Oh.

>> Go for it. (Fabian laughing)

>> I mean, we really see user using GPU, so we can expect that they want to run AI workloads. I don't think we have many feedback if, which kind of workload runs inside VMs, but yeah, the interesting GPUs is of course there.

>> Yeah, like the end user feedback, like sometimes are difficult, but I know for sure that like this is finding its way into KubeVirt, right? I mean, in the beginning there was like uncertainty, how does, plus KVM is old, right? I mean it's been used in OpenStack, it's been used in other projects out there for years. So what I want to say is, yes, we see AI workloads landing in VMs because there's some benefits to it as well. We see tech workstation use cases being new databases, everything. And you know, the nice thing is it's run as before, right? So you get the same stability, the same compatibility that, you know, have in other KVM based stacks. It's just that you manage differently, and we see that. I mean all like, just to conclude that like-

>> Yeah.

>> With all the preferences and instance types that we've been doing, like Windows, for example, it's like a common workload in VMs, and it's being used. We have complaints about not being able to automate as much, which is difficult with some operating systems, but that's a good sign, right? Because we see that those workloads get into KubeVirt.

>> Another interesting use case is deploying Kubernetes in Kubernetes. So KubeVirt is like the deployment to have maybe additional cluster, nested cluster, so this is also another perfect use case for virtualization.

>> Yeah, I think it's really interesting, actually I want to dive into the AI side of things. What sort of trends are you seeing in terms of how people are using KubeVirt to manage those workloads?

>> AI will rule the world.

>> Yeah. (Savannah and Rob laughing)

>> Savannah: Enough said sound bite, sound bite sorted.

>> KubeVirt can be two sided, right? I think AI is finding its way into, you know, into operator's daily life. Like, I mean AI is approaching us from all sides, right? It's helping operators, it's helping us with data, big data, right, to analyze it, all the large language models say about big data. But we also have it as a workload. So I think like we will see AI, you know, approaching KubeVirt from many sides, as a workload, but also from the Kubernetes side, like everything that's like AI and integration with Kubernetes, we are expecting it, you know, at least touch KubeVirt as well to some extent. But we need to see. I'm not aware of anything specific related to KubeVirt and AI at this moment in time.

>> Yeah.

>> Do you have any thoughts?

>> I think, as I said, the use of GPUs, really, I think it's the result of the interest in AI workload, and I mean API is one of the main contributors to KubeVirt. So definitely people are interested in using AI workload with the VMs.

>> I was going to say they were on stage yesterday talking about the APIs and things of that nature. So there are, is that a big direct input into what, how KubeVirt is using the GPUs and the hardware underneath?

>> So a big shout out to them. I mean they're really taking it really serious, they do a lot about scale and performance in KubeVirt. Yeah, they have a SIG-scale, right? They're driving it, that's great. And yeah, they optimize it for their use case, including GPUs.

>> Yeah, so how are things go from KubeVirt to being in OpenShift virtualization? And kind of what are the benefits of that? 'Cause I mean, again, you are here from Red Hat, so you can put your red hat on for a second, you know, (Rob laughing) and talk about that. How should people think about that and how close is it to KubeVirt?

>> So I mean, usually installing and managing Kubernetes is hard. So you can really think Kubernetes as Linux and the same to OpenShift, so actually, so OpenShift really gives you the entire package and installation. So, and we work very closely with the OpenShift team, so it's very well integrated. So when you do enterprise, you probably want to have a full solution, and this is the benefit that you get with OpenShift and KubeVirt and Cloud Native virtualization.

>> So it's just part of the whole package, then?-

>> Yeah, you can install it through the operator hub and it works like every add-ons on from OpenShift.

>> Yeah.

>> Sweet.

>> Yeah, like first it's very closed, right? I mean for sure, like KubeVirt is built like, I mean we build it is Red and chip it with OpenShift and it's tied into their use system because the reality is Kubernetes alone and KubeVirt alone, you seldomly see it, right? But it's with any other project, right, you need that ecosystem, right, you need observability, you need storage, you need networking. And OpenShift, after all, OpenShift by itself is an enterprise Kubernetes platform that is giving you like this boilerplate, right? To focus on what is bringing value to the user, right? About running the applications. And OpenShift virtualization is just benefiting from this, right? We have OpenShift as a solid platform and we add virtual machine support as part of OpenShift virtualization and it ties naturally into what the admin is already doing. Right, so they don't need to rescale it, so they can use what they know working with Prometheus, right, working with Loki, there's nothing they need to learn new except for working now with the VMs as well. But that's just another workload, nothing-

>> Just another workload.

>> Just another workload.

>> Love the whole ethos, you've had some great sound bites, both you today. Last question for you as we wrap up, what do you hope you can say at the next KubeCon that you can't say today?

>> Ooh, good questions.

>> The next KubeCon that we can't say today.

>> Yes. Where do you hope we're progressing, or the project's progressing or?

>> So definitely getting more contributors, that will be a great achievement.

>> I like that.

>> I think seeing that, I would, now switching forward to 2020, it's great to see that finally people realize that Kubernetes can be a, you know, it's a compute platform that can be run on bare metal as well. It's not only for application developers, it can also be a tool, and that's fantastic. Being on bare metal used to automate your work, I mean, containers are not just a tool to drive applications. We need to understand, as an administrator, I need to administer my infrastructure, and the term "operators" coming from the fact I'm codifying operational knowledge, right? So why can't we codify an administrator's knowledge to operate our infrastructure? So I would love to, I'm expecting, no, I'm looking forward to see that Kubernetes is recognized as such, right? As a compute platform, right? And that is running multiple workloads, so as a tool for administrators as well.

>> I love it, that was great. You're both absolutely fantastic, thank you for joining us this morning.

>> Thanks to you too.

>> Making the time in your busy schedules between keynotes and everything else. And Rob, thank you as always-

>> Thank you.

>> For your wonderful insights and questions. And thank all of you for tuning in to our three days of live coverage here at KubeCon CloudNativeCon, CNCF's biggest European event. My name's Savannah Peterson, here in Paris, France, you're watching "theCUBE", the leading source for enterprise tech news. (bright upbeat music)

Alice Frosi & Fabian Deutsch | Red Hat

search

(bright upbeat music)

>> Good morning, Cloud community, and welcome back to spectacular Paris, France. We're here at KubeCon, CloudNativeCon, CNCF's flagship European event. My name is Savannah Peterson, delighted to be joined by my fellow analyst, Rob Strechay. Rob, did you have any pastries this morning?

>> I did not, I did have a baguette though. (Savannah laughing)

>> You had a baguette this morning?

>> I had a baguette this morning, that was my breakfast, so.

>> I'm actually a little jealous, I might-

>> Yeah, it was fantastic.

>> You didn't want to bring a baguette for the rest of us?

>> I was on the walk. (Rob laughing)

>> Oh, it was on a walk. All right, all right, all right, well-

>> On the move.

>> At least I'm virtually having a baguette right now, 'cause we're talking about virtualization on this next panel. Very excited to welcome our two guests, Alice and Fabian. Thank you so much for being here, how you all doing?

>> All good, yeah, I'm glad to be here.

>> Savannah: Is this show a big event for you?

>> Yeah, sure-

>> Savannah: Yeah.

>> KubeCon is like the event. (Alice laughing)

>> Savannah: It is the event.

>> Yeah.

>> Yeah, we're also really big, super deep nerd fans. Fabian, you were on stage today.

>> Fabian: Yeah, indeed.

>> Tell us about that.

>> Tell us about-

>> Tell us about that. You're on this stage now, and we want the highlights.

>> It was, like, I feel, you know, it's great, right, to have that opportunity, and to have such a great co-speaker who can't be here right now, but like, it's amazing, right, so you-

>> Savannah: It is.

>> You have somebody, they've adopted KubeVirt, it solves a problem to them and then you can get into such a stage. I mean, it's like wonderful-

>> Rob: Well, it was Goldman Sachs who was on-

>> It was.

>> Talking about how they're using KubeVirt to actually move virtualized machines, or virtual machines, into their Kubernetes environment. Which I think is, I kind of talked about it yesterday, how I was surprised we hadn't talked about it, and I totally forgot you guys were coming on today. But now we get to talk about, but-

>> Or did you forget? Subliminally I think-

>> Maybe.

>> You were definitely teasing this interview the whole time.

>> I may may have been teasing this interview the whole time-

>> I love it.

>> I think again, it's such an important topic I think, and a project in general, because of the changes that are going on in the industry with Broadcom buying VMware, there's a lot of service providers, for instance, that are looking for other options for doing hosting of virtual machines and things of that nature because of the contract changes that have gone on underneath there. Do you see a surge of people being interested in KubeVirt since-

>> Good question.

>> All of this has started happening?

>> Sure, I mean, even here at KubeCon, we got many questions how people can transition from VMware to KubeVirt, if KubeVirt is stable enough to be used in production. So we have seen a lot of interest in the last month so.

>> And I bet you're not surprised by that.

>> I'm not, I'm not. (Savannah laughing) So what's new with KubeVirt? What's been going on?

>> So right now, KubeVirt is like reaching kind of maturity so we are focusing more in clean up the code, get more people involved. So there has been many initiative to improve the code base, to clarify more the ownership. There has been also a lot of improvement in the user guide, as many people are joining the community, this is of course the first entry point for the users.

>> As Alice just said, right, so we're working a lot, like, you know, with KubeVirt, we had a race, right? I mean, with KubeVirt, we want to provide, we are providing a virtualization solution based on Kubernetes, and just one addition to what you said about Goldman, like you know important is, you know, Kubernetes is becoming the infrastructure and VMs' become one workload that's being run there, right? And KubeVirt is enabling to do that. So we've been working with KubeVirt, right, to enable all these features that you expect from, you know, these virtualization systems that are out there today, and we've achieved much of that over the past few years. Like some big things like hotplug for CPU memory resources are still, they need to GA. Memory overcommit, higher density, that's something we're working on. But we're switching to, you know, being even more stable, right? We provide seamless upgrades, but we want to be more stable. We're working on, you know, making easy to maintain. Because if you switch KubeVirt, you can rely on it, right? And that's what we want and have to live up to, so we're changing focus a little bit, while not, I mean, while continuing to focus on innovation as well, right? New features have to go in, but the foundation needs to be solid.

>> Yeah, another interesting thing is usually virtual machine are hard, they have many options. So we have tried to simplify the deployment, so there is a new feature, and-

>> Savannah: Super important, everyone wants to decrease complexity.

>> Yeah.

>> Fabian: Yes.

>> Savannah: Yeah.

>> So for example, we got also many questions about Windows, how to deploy Windows VMs, so you can define, actually, if you want to run a Windows VM, there is like something called distance types, and this helps user to deploy a virtual machine. So this kind of template that would set all the option for you, so this, again, to make it easier for user to deploy VMs.

>> Everyone wants that easy button, we talk about it a lot-

>> Yeah.

>> And especially when it comes to Kubernetes, especially when we're talking about AI and everything else. You know, you and I were talking about your daughter's senioritis earlier-

>> Yeah.

>> And in thinking about graduation, where's the project maturity right now? Fabian, I'm going to go to you on that one.

>> Yeah, so, you know, I think we joined the CNCF incubator, in 2019, and incubation 2022. 2022, yeah. And that was always a good reflection of the project's maturity, I would say. So I think it's just, you know, the natural next step to now work on graduation, it totally aligns to what we're doing, like on a day-to-day basis. Like working stability, finishing the, as I just said before, right, so yes, graduation is definitely something we're working on this year. We're working actively like to reach out to end users, to more platform providers, like we have a lot of platform providers who are using Kubernetes and offering KubeVirt on top as a service or, you know, using it by themselves. So we're working with them to simply complete that whole graduation. It includes security reviews, it involves governance, and involves end users and other adopters.

>> How many contributors are on the project?

>> We have 150 contributors. (Alice laughing)

>> Alice: You've prepared today, so you-

>> Yeah.

>> Alice: You know the numbers. (Rob and Alice laughing)

>> And I'm even more about the companies, because the company ratio increased over time-

>> Savannah: Yeah.

>> It's always good for a project, right, to not rest on, you know, few companies. So it's great to see that more companies become interested, regularly contribute to KubeVirt, so that's just great to see.

>> Do you see people, what's I guess the ratio between people who are vendors contributing, versus non vendors that are contributing to this, do you think?

>> I would like to take that offline and work with you.

>> Rob: Okay.

>> To digest the data. (Rob laughing)

>> Rob: That sounds good.

>> I think like, you know, KubeVirt is for sure very interesting to platform providers. So vendors, right, you could say, and they're for sure, I would say they're more dominant, right-

>> Alice: Yeah.

>> In our contributors.

>> Alice: Definitely.

>> Rob: Yeah.

>> But some end users also contributing, right?

>> Yeah. And where are kind of the requirements coming from? Is it people saying, "Hey, this is what we've seen in KVM," "this is what we've seen in VMware," "these are the things that are missing." How is that working with the community?

>> So I mean, of course people coming from traditional virtualization platform, they expect certain feature to work, like for example, hotplug, as Fabian mentioned. But we are also getting requirement like from container perspective, like, I dunno, like selecting certain environmental variables or passing certain kind of files. So I would say it's kind of a mixture, but yeah, a big part of the requirement is for sure coming from virtualization, the virtualization world.

>> And it's leveraging, one of the other keynotes this morning, they were talking about the storage aspect and it was, it's leveraging CSI to be the connection back to storage. And is that, are you being able to leverage a lot of the different things that have been done for, you know, things more ephemeral container types, but for something that needs to be more persistent?

>> Yes, definitely. So we, KubeVirt doesn't make any difference. So between storage, storage provider, so if you have a CSI plugin, it can work with the file system, and it can also work with block devices. So yeah, generally if you have your CSI plugin, it can perfectly integrate with KubeVirt.

>> And can I just add something?-

>> Please.

>> I think that's true for so many aspects, right? I mean, why did we build KubeVirt and Kubernetes at all, right? Because Kubernetes is providing so much, right. From an infrastructure perspective, CSI, CNI, but also all these other tools like, you know, patterns for observability, patterns for our deck, right? Policy enforcements and such, that stuff, so that's why we built KubeVirt on top of Kubernetes, to leverage it. Why should we reinvent it again for VMs? How can we build such a wonderful ecosystem that we have with the CNCF ecosystem?

>> Yeah.

>> We could not build that for VMs only, right? So like by embedding VMs into Kubernetes, we can simply benefit from that ecosystem that's providing so much, it's like.

>> I mean, it makes sense.

>> Yes, yes.

>> Everything makes sense. So what's next? What are you going to build next? What are the contributors going to make next? What sort of features can we expect?

>> What's on your roadmap for this year-

>> Yeah, exactly, yeah, give us the preview.

>> So live update of resources is of course a big topic, so we just added a CPU and memory hotplug. So we were already supporting storage and network interfaces, so we can expect also, in the next future, to have additional feature in this direction.

>> Do you see different workloads coming to KubeVirt that are non-traditional VM workloads? Like some of the AI stuff, some of the database stuff would make sense to me, which is not a, I know you can do persistent databases on, and I sat in a Postgres persistent database talk, data on Kubernetes day on Tuesday. Do you see those types of workloads really being driven to KubeVirt as well?

>> Oh.

>> Go for it. (Fabian laughing)

>> I mean, we really see user using GPU, so we can expect that they want to run AI workloads. I don't think we have many feedback if, which kind of workload runs inside VMs, but yeah, the interesting GPUs is of course there.

>> Yeah, like the end user feedback, like sometimes are difficult, but I know for sure that like this is finding its way into KubeVirt, right? I mean, in the beginning there was like uncertainty, how does, plus KVM is old, right? I mean it's been used in OpenStack, it's been used in other projects out there for years. So what I want to say is, yes, we see AI workloads landing in VMs because there's some benefits to it as well. We see tech workstation use cases being new databases, everything. And you know, the nice thing is it's run as before, right? So you get the same stability, the same compatibility that, you know, have in other KVM based stacks. It's just that you manage differently, and we see that. I mean all like, just to conclude that like-

>> Yeah.

>> With all the preferences and instance types that we've been doing, like Windows, for example, it's like a common workload in VMs, and it's being used. We have complaints about not being able to automate as much, which is difficult with some operating systems, but that's a good sign, right? Because we see that those workloads get into KubeVirt.

>> Another interesting use case is deploying Kubernetes in Kubernetes. So KubeVirt is like the deployment to have maybe additional cluster, nested cluster, so this is also another perfect use case for virtualization.

>> Yeah, I think it's really interesting, actually I want to dive into the AI side of things. What sort of trends are you seeing in terms of how people are using KubeVirt to manage those workloads?

>> AI will rule the world.

>> Yeah. (Savannah and Rob laughing)

>> Savannah: Enough said sound bite, sound bite sorted.

>> KubeVirt can be two sided, right? I think AI is finding its way into, you know, into operator's daily life. Like, I mean AI is approaching us from all sides, right? It's helping operators, it's helping us with data, big data, right, to analyze it, all the large language models say about big data. But we also have it as a workload. So I think like we will see AI, you know, approaching KubeVirt from many sides, as a workload, but also from the Kubernetes side, like everything that's like AI and integration with Kubernetes, we are expecting it, you know, at least touch KubeVirt as well to some extent. But we need to see. I'm not aware of anything specific related to KubeVirt and AI at this moment in time.

>> Yeah.

>> Do you have any thoughts?

>> I think, as I said, the use of GPUs, really, I think it's the result of the interest in AI workload, and I mean API is one of the main contributors to KubeVirt. So definitely people are interested in using AI workload with the VMs.

>> I was going to say they were on stage yesterday talking about the APIs and things of that nature. So there are, is that a big direct input into what, how KubeVirt is using the GPUs and the hardware underneath?

>> So a big shout out to them. I mean they're really taking it really serious, they do a lot about scale and performance in KubeVirt. Yeah, they have a SIG-scale, right? They're driving it, that's great. And yeah, they optimize it for their use case, including GPUs.

>> Yeah, so how are things go from KubeVirt to being in OpenShift virtualization? And kind of what are the benefits of that? 'Cause I mean, again, you are here from Red Hat, so you can put your red hat on for a second, you know, (Rob laughing) and talk about that. How should people think about that and how close is it to KubeVirt?

>> So I mean, usually installing and managing Kubernetes is hard. So you can really think Kubernetes as Linux and the same to OpenShift, so actually, so OpenShift really gives you the entire package and installation. So, and we work very closely with the OpenShift team, so it's very well integrated. So when you do enterprise, you probably want to have a full solution, and this is the benefit that you get with OpenShift and KubeVirt and Cloud Native virtualization.

>> So it's just part of the whole package, then?-

>> Yeah, you can install it through the operator hub and it works like every add-ons on from OpenShift.

>> Yeah.

>> Sweet.

>> Yeah, like first it's very closed, right? I mean for sure, like KubeVirt is built like, I mean we build it is Red and chip it with OpenShift and it's tied into their use system because the reality is Kubernetes alone and KubeVirt alone, you seldomly see it, right? But it's with any other project, right, you need that ecosystem, right, you need observability, you need storage, you need networking. And OpenShift, after all, OpenShift by itself is an enterprise Kubernetes platform that is giving you like this boilerplate, right? To focus on what is bringing value to the user, right? About running the applications. And OpenShift virtualization is just benefiting from this, right? We have OpenShift as a solid platform and we add virtual machine support as part of OpenShift virtualization and it ties naturally into what the admin is already doing. Right, so they don't need to rescale it, so they can use what they know working with Prometheus, right, working with Loki, there's nothing they need to learn new except for working now with the VMs as well. But that's just another workload, nothing-

>> Just another workload.

>> Just another workload.

>> Love the whole ethos, you've had some great sound bites, both you today. Last question for you as we wrap up, what do you hope you can say at the next KubeCon that you can't say today?

>> Ooh, good questions.

>> The next KubeCon that we can't say today.

>> Yes. Where do you hope we're progressing, or the project's progressing or?

>> So definitely getting more contributors, that will be a great achievement.

>> I like that.

>> I think seeing that, I would, now switching forward to 2020, it's great to see that finally people realize that Kubernetes can be a, you know, it's a compute platform that can be run on bare metal as well. It's not only for application developers, it can also be a tool, and that's fantastic. Being on bare metal used to automate your work, I mean, containers are not just a tool to drive applications. We need to understand, as an administrator, I need to administer my infrastructure, and the term "operators" coming from the fact I'm codifying operational knowledge, right? So why can't we codify an administrator's knowledge to operate our infrastructure? So I would love to, I'm expecting, no, I'm looking forward to see that Kubernetes is recognized as such, right? As a compute platform, right? And that is running multiple workloads, so as a tool for administrators as well.

>> I love it, that was great. You're both absolutely fantastic, thank you for joining us this morning.

>> Thanks to you too.

>> Making the time in your busy schedules between keynotes and everything else. And Rob, thank you as always-

>> Thank you.

>> For your wonderful insights and questions. And thank all of you for tuning in to our three days of live coverage here at KubeCon CloudNativeCon, CNCF's biggest European event. My name's Savannah Peterson, here in Paris, France, you're watching "theCUBE", the leading source for enterprise tech news. (bright upbeat music)