The player is loading ...
No more VMs? Kubernetes Multi-Tenancy with Loft Labs | Episode #98
No more VMs? Kubernetes Multi-Tenancy with Loft Labs | Episode #98

In Episode 98 of Great Things with Great Tech, Lucas Gentele, CEO of Loft Labs, shares how Loft is reshaping infrastructure by eliminating the need for virtual machines. Starting from early entrepreneurial roots and open-source projects like DevSpace, Lucas explains how Loft Labs tackled Kubernetes inefficiencies by building multi-tenant solutions like vCluster and vNode. Their innovations enable Kubernetes clusters to run more securely, efficiently, and cost-effectively without traditional VM overhead. As companies seek alternatives to legacy virtualization, Loft Labs mission is to bring lightweight, scalable, and open-source virtual clusters into more environments, making Kubernetes accessible for both development and production workloads

Apple Podcasts podcast player badge
Spotify podcast player badge
Castro podcast player badge
RSS Feed podcast player badge
Breaker podcast player badge
PocketCasts podcast player badge
YouTube Channel podcast player badge
Apple Podcasts podcast player icon Spotify podcast player icon Castro podcast player icon RSS Feed podcast player icon Breaker podcast player icon PocketCasts podcast player icon YouTube Channel podcast player icon
Season 2025 Kubernetes Infrastructure Platform Hosting Multi-Tenancy

We all know that virtual machines have dominated IT infrastructure for 25 years... but is that era really at an end?In Episode 98 of Great Things with Great Tech, Lucas Gentele, CEO of Loft Labs, shares how Loft is reshaping infrastructure by eliminating the need for virtual machines. Starting from early entrepreneurial roots and open-source projects like DevSpace, Lucas explains how Loft Labs tackled Kubernetes inefficiencies by building multi-tenant solutions like vCluster and vNode. Their innovations enable Kubernetes clusters to run more securely, efficiently, and cost-effectively without traditional VM overhead. As companies seek alternatives to legacy virtualization, Loft Labs mission is to bring lightweight, scalable, and open-source virtual clusters into more environments, making Kubernetes accessible for both development and production workloads

Key Takeaways:

  • The journey from DevSpace to founding Loft Labs and the rise of Kubernetes multi-tenancy.
  • Why the traditional VM-based model is becoming obsolete — and what’s replacing it.
  • How Loft’s vCluster and vNode are revolutionizing Kubernetes cluster management.
  • The impact of Broadcom’s VMware acquisition on cloud infrastructure strategies.
  • The future of cloud-native infrastructure: virtual clusters, no VMs, better scalability.

Links & Resources: 

Loft Labs: https://www.loft.shLukas Gentele on LinkedIn: https://www.linkedin.com/in/genteleLoft Labs on LinkedIn: https://www.linkedin.com/company/loftlabs

☑️ Support the Channel: ⁠⁠⁠⁠https://ko-fi.com/gtwgt⁠⁠⁠⁠

☑️ Be on #GTwGT: Contact via Twitter @GTwGTPodcast or ⁠⁠⁠visit https://www.gtwgt.com⁠⁠⁠

☑️ Subscribe to YouTube: ⁠⁠⁠https://www.youtube.com/@GTwGTPodcast?sub_confirmation=1⁠⁠⁠

Check out the full episode on our platforms:

Spotify: ⁠⁠⁠https://open.spotify.com/episode/2l9aZpvwhWcdmL0lErpUHC?si=x3YOQw_4Sp-vtdjyroMk3Q⁠⁠⁠

Apple Podcasts: ⁠⁠⁠https://podcasts.apple.com/us/podcast/darknet-diaries-with-jack-rhysider-episode-83/id1519439787?i=1000654665731⁠⁠⁠

Follow Us:

Website: https://gtwgt.com

Twitter: https://twitter.com/GTwGTPodcast

Instagram: https://instagram.com/GTwGTPodcast

☑️ Music: https://www.bensound.com

  • Virtual machines have been the cornerstone of IT infrastructure for the best part of the last 25 years. But imagine a world where virtual machines are no longer needed, no longer running applications and not even part of the general conversation. On episode 98 of Great Things with Great Tech, I'm speaking to Lucas Gental, CEO and co-founder of Loft Labs, and their mission is simple.
  • to provide infrastructure without the need for virtual machines. Making Kubernetes multi-tenant, secure, efficient, and cost-effective by building tools that enhance developer productivity, reduce infrastructure waste, and simplify complex cluster management. This is episode 98 of Great Things with Great Tech with Lucas Dante of Loth Labs.
  • Hey Lucas, welcome to episode 98 of Great Things with Great Tech. It's great to have you here. Yeah, thanks for having me. Awesome. Hey Lucas, I've been looking forward to diving back into the world of Kubernetes like we like we talked about in the pre-show just before we got our stuff together and what we're going to talk about. It's been a while.
  • Um, I feel like Kubernetes has not dropped off the face of the earth, but I think from a in a good way, it's not become like it's not that this hype cycle part of the of where it was back in two or three years ago, right? Which means in theory, it's gained much more mainstream um sort of adoption. So, we'll get into that a little bit later on, but firstly, I'm interested in yourself.
  • Um, you know, how did you get started in computing and IT and how did you end up, you know, getting to the point where you were founding Loft Labs? Yeah, I started programming uh when I was like I don't know 12, 13, something like that, right? And uh you know, built my own like little website and then started building our school website uh back in high school, you know, uh or I guess middle school kind of time and um you know, started my own one person little company building websites for you know, the local dentists and lawyers and and whatever was town, right? Uh I think
  • I always had that entrepreneur that kind of spirit. I always had a curiosity for for just kind of building things and diving into things and um I think I was maybe a little naive in the sense of like oh yeah yeah I can you know create a business no problem right um and uh yeah I think uh in college I did the same thing I hired every good uh engineer I met there and and we started doing contracted uh work and that was was a lot of learning and a lot of fun.
  • Yeah, I mean it's it's funny if if you think about a lot of a lot of people that started when they were younger or you know got into computers. I mean I know that I certainly hacked together a few websites for for people when I was out of high school and whatever and you know managed to charge a little bit of money for them here and there.
  • I mean I it never really took off but you know I kind of enjoyed that side of it because people were after that skill set and it all came from a bit of dabbling and a bit of tinkering and kind of doing your own thing and that's kind of how it happens right. So is that something that you always had? you always had that desire to to code yourself, but I guess what I'm hearing is you did that, but then you realized pretty quickly if I get other people to do it for me, then I'm going to be more successful at that.
  • Yeah. I mean, I think in college it wasn't quite as handsoff as it may sound, right? It was a was pretty much a team effort. But, uh, yeah, I I definitely realized, okay, how do I go from amplifying my myself, right? Like, and and the easiest way is to form a team around you, right? And then build things with five people, right? rather than uh on your own.
  • And I guess today you know building the company I'm seeing the same thing. You know we're 60 people now it's kind of wild but uh you know you know 12 months ago we are with half the amount of folks and today you know if you ask me about especially some you know specific roles I'm like I don't even know how we did this before right it's kind of wild. Yeah.
  • So everything that you done before kind of is a learning journey to get you to the stage that you're at. I mean what so what were you doing out of so when you're at uni when you finished uni because obviously you know this is this isn't sort of the last couple of years what what did you start off with how where was your first gig in this industry after you know the tinkering and college yeah it's actually quite interesting because uh through our college like I I was actually working at the the chair of artificial intelligence uh back then um
  • and you know they there was this opportunity to take like a year of uh it's kind of a research entrepreneurial kind of year um and and the government kind of, you know, uh provides you with a salary for a year and you can Okay. Uh you know, it's it's not a pretty great salary, right? Like folks that went off to to work for Mckenzie and all of these consultants who are trying to hire us um you know, they they obviously had you know 3x uh the amount of of cash available, but it allowed you to uh you know there were no strings attached in
  • in that sense, right? It was really like okay explore what you want to do whether you want to become an entrepreneur and you know uh start your own startup and you you they had these educational pieces where other folks in this program across the whole country kind of came together and exchange ideas and you know there were coaches talking to you about you know um how how would you start and it was very interesting uh for me to uh to dive into that world because I've built this you know uh more of um um services kind of business before rather
  • than actual startup not building a product not raising funding. It's a very different journey if you are in this uh contractor service based world rather than uh the product focused world and so there was a lot of learnings for me I mean I didn't know what why combinator was you know at the start of that journey even right and that comes up later I believe as well why combinator comes up later as part of this journey right yeah I understand yeah we we built this open source project called devspace it's still around still a lot of people
  • are using it CNCF sandbox project now we handed it over to to the to the CNCF. Um, this project was the starting point. We built it for ourselves and it was our first tool in the Kubernetes space because we we started working for a lot of companies that were running on top of Kubernetes.
  • We wanted to build software on top of Kubernetes and okay the development flow was just rather painful with Kubernetes. So um we had to come up with a better way and and Devspace was was our better way essentially and So what is Devspace? What is it? Like this is 2018. This is like and Kubernetes has only been around what for like what three four years really and and really know the majority Kubernetes journey and honestly the majority don't even know what Kubernetes is.
  • Let's be honest like it's just kind of starting. So what was Devspace? Yeah it's essentially a developer tool for Kubernetes. It is a if you will docker compose but for Kubernetes. So, you know, docker compose is this great tool. You run one compose up command. You have a compose file in your in your repository, right? And you fire up multiple containers.
  • They're interlin to each other. The developer experience is pretty smooth because you're hooking your uh local laptop volume into uh one of the containers um you know, volume mounts, right? So you can essentially rebuild that container really really quickly by just kind of like having a file watcher rebuilding inside the container and that way you can iterate super quickly.
  • With Kubernetes that becomes a little bit more challenging because you know your Kubernetes cluster may not even run on your local machine. it may run um in AWS and you want to integrate it with uh you know services uh like RDS for example and you want to use IM authentication like these kinds of things like there's a lot of complexity when it comes to uh AWS and and Kubernetes and and other cloud providers right so you know we built Devspace we kind of modeled after docker compose but designed for Kubernetes we had this file syncing mechanism to hot
  • reload containers effectively and to do live debugging inside a container. Uh so the goal was really like it feels like local programming with with Docker Compose and a local Docker demon but it all happens inside a Kubernetes cluster that may be local or maybe remote, right? That was that was the goal. Okay.
  • And then this is like and this is so do you have did you have some sort of funding for this? Was this a was this the Y cominator scenario where you were living there and doing sort of that startup thing? But but you weren't but you weren't like that straight out of college situation though, right? So how how did how did that go? Because you're probably I mean you're not that old but you were a little bit more mature potentially than other people in that sort of space.
  • Did they make you like live like a startup or what was the deal? I I think uh yeah so we we applied for Y Combinator at that time. We had this open source project. It's gotten traction, right? There was a a lot of attention and we got hooked on open source as well, right? That was the first time us not contributing a line here or there like but us maintaining the project, right? We are suddenly responsible for it.
  • People asked about feature requests, right? We we are you know pushing the road map forward. It was a great experience and you know we really fell in love with open source and we had this idea of okay what if we built like a pass uh similar to like a heroku based on that dev space experience. So we applied for Y cominator and Y cominator was really excited about it.
  • uh you know they invited us at this final round of interviews. They flew us to California the week after, you know, a Skype interview and um ultimately they told us, hey, it may be a little bit premature and you have some questions to to answer in terms of, you know, where this where this should go and they were skeptical about passes in general, I felt like and they ended up being right because obviously we pivoted and and built something different.
  • So they sent us back home and said like why don't you apply again in six months? We do this twice a year, okay? uh and and you think about these kinds of questions, we'd love to see you again. That was essentially the feedback there. And at the same time, uh you know, we submitted for a couple of these startup competitions, business plan competitions, uh back in Germany.
  • And uh you know one of these competitions the part of the the the winning prize we had there was you can apply for UC Berkeley's Sky Accelerator program without uh being affiliated with UC Berkeley because you have to be an alumnest at that school in order to apply. I think I think they lifted that requirement uh recently but at the time that was still a requirement and you know neither of us uh was at uh UC Berkeley at the time so obviously that wasn't an option but through that uh special you know grant there we we got
  • in there and they they said like okay cominator was excited about you guys but they told you to come back in six months we fund you right now pretty smart about about this you know we can steal these founders away before they before they end up at YC and it's the same experience, right? They bring it.
  • Yeah, I know. We we tried to make uh dev space work. Um but ultimately building that pass was really difficult. I think building a pass ultimately you're a hoster. We didn't realize that as much but you're literally competing with AWS and you know there's also a lot of other hosting companies that are trying to compete with AWS on a cheaper price.
  • So the past space is like nobody is appreciating that software layer on top, right? So that was quite challenging but what it taught us you know we had a lot of users in the free tier not a lot of paying customers right that's the way we were forced to build Kubernetes in a very very cheap and efficient way and multi-tenency was the solution for us there was no way we could have handed a Kubernetes cluster to every individual user but name spaces weren't good enough for a lot of scenarios so we had to overcome that and you can kind of see
  • where this is heading this is this is our journey towards lending vlo start. Yeah, because this is the thing, right? This is the whole I guess this is the whole value prop. This is the whole reason for being this is the problem statement that you're trying to trying to trying to solve with Loft Labs is that Kubernetes isn't really very friendly to multiclustering and shared tenency.
  • It's it's just not like I know trying to play with it in the early days and trying to sort of wrap my head around it, different flavors, whatever it was. And I think I mentioned earlier in the pre-show that VMware came the closest I felt to some semblance of multi-tenency but even that was super clunky with the supervisors and how that did it and but that wasn't scalable because it wasn't cost effective as well and it still leveraged vSphere and everything like that.
  • So this is the the reason why Loft Labs exists is because you were working through that process but you had to come up with a way to to make everything cheaper and you thought why don't we just why don't we just virtualize this thing and make it multi-tenency. So how how does that come about? So how do you without giving away the trade secrets how do you make you know Kubernetes multi-tenant with vCluster? Yeah, I think the the core idea is really to uh put a Kubernetes control plane inside a container but remove the scheduleuler and you know this is all
  • part of the open source vcluster. So there there's really no trade secret in that sense in there. You can look at the code and how we did it. Uh we removed theuler from the control plane because what auler needs is nodes and as soon as you need nodes you have a completely separate cluster because how do you create a node? Well, that can be a virtual machine or it can be a physical node.
  • All of that would essentially mean you you kind of have a separate cluster, right? But when you want to do multi-tenency, you want to have a set of 400 nodes and then all of your tenants should somewhat share these nodes and you know maybe some tenants get a you know dedicated node for specific things, right? They need specific privileges in their you know root containers, host path mapping, these kinds of things.
  • But for your average you know web server or average type of application in Kubernetes where that interacts with message cues etc it's totally fine to run on a shared node if you you know isolate well and secure it. Um so the idea was really let's remove auler. So because one node can only be in one cluster at a time.
  • So one, right? If we remove theuler and instead replace it, what we call syner, we can copy parts from the virtual space. The virtual cluster has its own state. It has its own control plane, separate API, ser controller manager, separate CD, but can also be a SQL or my SQL or any kind of backend store that we're supporting.
  • It's a whole bunch of options. Then you know you create a deployment replica number two that creates two pods. Everything so far is just an entry in your data store. Nothing actually happens in terms of spinning up a process, right? Spinning up a container. But now theuler is supposed to take that part and put it on a node.
  • Instead, what we do, the syncer copies the part to the underlying, you know, real physical traditional Kubernetes cluster and then thatuler in this cluster is going to schedule this part on one of the nodes that is owned by that bigger cluster that's underlined. We call it host cluster. That's the rough magic of VCluster. Obviously, there's a lot more to it in terms of like preventing naming conflicts and isolating these workloads and network policies and all of this kind of stuff to make it to make it resilient, right? But this is this is
  • the essence of it, right? Let's get rid of theuler and and put this in. Yeah. So in a way you're you're sort of virtualizing you know Kubernetes to a certain extent but then you've got this goes a whole step further with your v node which we'll get on to later but this is effectively what you're doing you're kind of virtualizing the name spaces and and creating enough wrappers around it and enough conditions that the user can log in and have their own multi-tenant component and that in itself just allows like you said it
  • allows cost efficiencies right and where where is this used so what what were your typical early day customers what were they looking to achieve with using vcluster I think it it evolved over time initially when you launch thing like vcluster it's a credibility question because we're going out saying how you guys are doing at this fortune 500 kubernetes cluster is the wrong architecture a lot of people at the time and still today have duty and spinning up virtual clusters so that you hand out these virtual clusters instead of you know separate
  • clusters. So initially we saw a lot of traction with pre-production workloads because folks were like okay this is a new technology this is a new architecture we don't trust this in production quite yet it's just of containers and containers came out there it's not like you know the biggest banks in the world immediately said like okay let's move our mainframe stuff into containers right yeah it took quite some time for people to gain that confidence and the same happened with vosta people used it a lot as sandbox environments
  • development environments, learning environments, right? Sometimes they ran workshops with the cluster for example, which I always thought is a really fascinating use case. We never thought of that when we actually launched. I was thinking that just then like I think you know in terms of what I what we were doing we were um at veh we had casten right our kubernetes backup solution and there were little um sort of live labs and and workshops that we were running and you know that's perfect use case for that sort of live lab where you want to
  • spin something up really quickly and just kind of have it ephemeral and and shut it down but have multiple people accessing it without it costing so much money. So you're right, it's it's out of necessity that these sort of things come in. But then to your point, um, use cases become clearer in enterprise and commercial senses and that's where you're at today.
  • Yeah, absolutely. I think people have understood the vCer architecture. they have more confidence in Kubernetes multi-tenency and they go beyond you know CI and dev use cases and now embarking on okay how do we run our production Kubernetes environment in a multi-tenant way and I think one of the biggest use cases we're seeing recently is going from offboarding uh of VMware offboarding of you know virtual machine-based Kubernetes clusters and moving towards bare metal clusters Yeah, because you're removing that
  • entire overhead of the virtual machine layer. And the only reason why you had that in the first place is so you could divvy up your environments into multiple nodes and multiple clusters, right? But with bare metal, you can have, you know, 500 bare metal nodes in one data center as one big kubernetes cluster.
  • then you can have virtual clusters on top instead of creating you know VMs and then creating these little clusters that are highly underutilized. We see a lot of that happening right now and the only missing piece besides you know control plane isolation that you get with vluster is how do you keep things separate on the node especially when some of your workloads may need you know there's noisy neighbor problems there's requests for host paths and there's requests for uh a privilege container container that runs as root and that's
  • highly problematic if you're running in a multi-tenant node that's why we announced the node at CubeCon and that's a very powerful layer that you put on the node to essentially create these virtual nodes and wrap your containers in these virtual nodes to still make them secure without the need to create static virtual machines. Yeah.
  • So this is last week at KubeCon when this is recorded at CubeCon which is I guess it's the the European instance of KubeCon for 2025 announcement there of of Vnode um which effectively is trying to eliminate VMs um get rid of get rid of VMs right this is this is your kind of motto you were telling me earlier which is a very bold statement to make right let's get rid of VMs let's get rid of the way that everyone's been doing stuff majority for the last 25 years but the timing is right at the same time we all understand the complexity of the
  • world that we're in today with a Broadcom and what they've done and how they've they've forced people into making them look into alternatives, right? So, if you're a an enterprise or a commercial entity that's been running VMware and all of a sudden you get stuck with what's happening there, you're going to look at alternatives.
  • Typically, the alternatives are um you know alternate hypervisors for sure with VMs just moving and doing the migration. Um and then what we've also seen in the Kubernetes world is is coupube specifically, right? But this isn't that. So just make the differential between what Vnode is versus what Kubvert is versus what a VM is.
  • Yeah, we've heard so many of our customers obviously we we work with a ton of companies that have put VC cluster to use and that you know obviously manage a larger state of Kubernetes clusters and they've tried several solutions right there's things like cubeword as you mentioned uh which is essentially just different virtual machines than the ones you get from VMware and the downside is it's really inferior technology because VMware really built a very efficient very lean being very quick to start up, right?
  • like there's so many benefits around that VM EOS VMware ecosystem that it's really really hard to move to a technology like cute word um which is much more immature in comparison and may never get to where you know VMware ended up being after you know 25 years or so right like a lot of innovation and R&D went into it product but you know all of these alternatives that are VMware B or VM based they have the downside side of VMs are still pretty heavyweight.
  • Even the ones from VMware are relatively heavyweight. Separate kernels. Everything's completely separate. VMs are relatively static, right? Um and when you look at containers, we we were hoping to turn, you know, these pet VMs into C into cattle and have dynamic containers on top of that. That was the whole promise of Kubernetes, larger clusters.
  • You know, at Google, there wasn't there wasn't 500 BS. there was one board that had all of these machines, right? And then the dynamic uh starting of containers that that is the benefit and allocation of resources dynamically. That's the big benefit that Kubernetes was initially promising and then we started creating these little three to five node clusters that are highly underutilized and now we have pet clusters instead of pet VMs, right? Yes.
  • Yes. That's essentially what we're trying to fix with VNODE. We're taking the more Kubernetes native approach. um folks have tried to you know move to bare metal kubernetes we've seen that in our customer base and they've used vcluster in combination with other technologies I'll give you a couple of examples right like cutter containers which creates microvms again still VMs not ideal people have used gisor which is seccom filtering and there's a whole bunch of these other tools right that try out alternative ways um what we're
  • using is username spaces in combination with seccom filtering So we don't give you an entirely separate kernel. It's still a container. That virtual node is still made up of a container, but it is as secure as you can get without actually creating a virtual machine. So I think there let's say for example the drivers are shared, right? If you would be using specific drivers and there's a driver vulnerability, that's potentially one of the only cases where VNO is a little bit less secure than a VM, but it comes at all of the benefits of starting
  • up super quickly, being very dynamic, no static resource allocation, um, and much much less overhead. We recently, you know, had a story uh going out there with a with a customer of ours called Aussie Broadband in Australia. The telco know them well. They're they're running this podcast right now. They're streaming this podcast.
  • That's awesome to hear. They went from 200 VMs on VMware to diameter Kubernetes zero VMs and they are launching virtual clusters on top and initially they were using G visor on the on the nodes right which has all kind of limitations we've seen that with multiple of our other customers it's very trivial even with you know G visor I think doesn't support even port forwarding and has all these intricacies it's not using user name spaces only secum filtering performance is is force than than what you see with vnode. So in these kind of scenarios uh
  • using vuster and vnode is really powerful and what a broadband did from 200 VMs to zero doesn't just save you the VMware licensing cost it also freed up a thousand CPU cores and a terabyte of memory just from the overhead of these virtual machines. Yeah. Which I can't comprehend that. That's huge. It's it's it's huge.
  • It's it's so significant. But you know, you mentioned obviously the limitations of of the coups and and the G-Syncs and everything like that and and it's a fair question to ask I think like what what do they give up like if they're moving from VMware which you've we know like we've I've been a VMware admin for a long time I I loved it built career out of it love the tech it was it's very featurerich like you said so are they were they what do they lose and are they running my question there is What were they running? Are they running running sort
  • of Windows boxes applications on them? Heavy heavy enterprise apps or were they running Linux machines that were a little bit more lightweight maybe? Like what was the makeup of that 200 VMs that they could shift it so easily um to end up with you guys at VNODE? Yeah, I think one part that is really important to understand is we don't help you with your VMware state that has legacy applications that are not containerized.
  • Right? So most companies out there, they may have let's say 500 VMs running legacy applications that are not containerist. We are not going to help you with that. Okay, that's really good. That's that's that's a really good that's absolutely a really good clarification to make for sure. That's great to hear. Yeah.
  • Yeah, we help you with the applications that are already containerized and where you had to spin up VMs to form separate Kubernetes clusters because yeah, people can't just have one cluster and run everything on one cluster. We just recently spoke to um they're not a customer yet. They are in the PC type stages with us.
  • Um I can't name any names, but it's a really fascinating story. They had over 5,000 VMs only for the Kubernetes estate and they started moving things over to a single big diameter cluster and they gave their applications individual name spaces only. The reason why they still have 5,000 of these VMs is these uh Kubernetes clusters that are running on top of these VMs and the workloads inside require to have the entire cluster do things across multiple namespaces have CRDs in Kubernetes all these things that don't fit in a
  • namespace. So they came to us essentially asking, okay, what if we were to now spin up VC cluster inside our bare metal cluster? Could we move these 5,000 VMs over quicker, right? These applications that are running in there because now it's moving one cluster to another cluster rather than a cluster and trying to mash it into an individual single namespace in Kubernetes and that's a huge game changer for them.
  • And then there's questions like how do we isolate things on the node level? How to give additional privileges to containers that need to run as root, right? And vnode is the answer for that piece as well. Okay, get it. Yeah. So, yeah. So, that that's it's good to it's good to make that demarcation, understand the use case there.
  • Um, but certainly, you know, we know that as applications modernize and net new applications get created, they're more than likely going to be cloud native in some way, right? Whether it's or whether it's, you know, from a container perspective as well, it seems just to be the natural way. So this is really banking on the future a little bit in terms of where you know the modern application is going or there'll always be legacy right but what you're saying here is if you've got a cloudnative application built in a modern way there's no point running it
  • on a VM I think that's when you say get rid of VMs that's that's really where where you're sticking that flag right the true cloud native Kubernetes way container way right that's what we're advocating for why are we even using that VM that is the question virtualization with virtual machines. I think that's that's probably pinning it down.
  • And when you when you're looking at the most cutting edge things going on in the data center right now is probably GPUs and all of these, you know, GPU cloud providers rivaling AWS, right? Which is larve had a couple on the show. Yeah, I've had a couple on the show. It's crazy, you know, but we would used to call them high performance compute providers back in the day, but now they're all kind of rebranded GPU providers, you know, and soon they'll be LLM LLM hosting providers basically, I feel as well.
  • But yeah, this and this is this is the point, right? I think this is a world where you need as much efficiency as possible because what you're trying to do is is abstract and slice that GPU, that time that the the transactions per second is what counts here, right? So yeah, and they don't even think about virtual machines, right? That that there's the world of folks that have a lot of virtual machines and build Kubernetes clusters with it and they want to move away from it.
  • And there's these these folks that have never even thought about virtualization because that 30 40% latency and that overhead is way too much for that expensive piece of hardware that they're that they're selling. That reminds me of a funny story. I was coming back from an AWS reinvent probably I think it was even maybe just before the pandemic or just after but I was sitting sat next to someone on the plane on the way back I was talking about what we do at veh um just general conversation trying to make some convo I sat and he said he never heard of us and I'm like well that's
  • okay we're biggest backup vendor in the world we back up virtual machines and he goes oh what's what's and what's what's VMware he asked me that he said I've got and it blew my mind that that there is this group of people that just don't know about to your point or care about virtualization the VM the VM world.
  • So you know let's be honest with you like there's a whole other world out there which you know if you're not if you're not crossed over like you are like I am and others you wouldn't know about. Um but certainly there's a group of people that are just there to do a particular job develop a certain way leverage GPUs or leverage something in in such a way they don't care whether it's a VM whether it's a container whether it's a bare metal they just want to get the job done.
  • So, you know, what you're enabling here is someone to just get the job done more efficiently, right? That's that's the way that I'm reading this. Absolutely. Yeah. Maybe we'll end up building that pass after all on top. Yeah. Go back to Y Cominator. Yeah. You can go back and say, "Hey guys, I told you this was going to be big. Just took us a while to get there.
  • " But no, no, no. I think I think that's really awesome. I think that's that's that it's a great the reason for why, you know, Loft Labs exist is is quite clear. And I wouldn't say it happened, it doesn't seem to me like it happened by accidents. It happened out of necessity, which I most of the uh companies I have on the show, this is actually how they get to their first product because they were doing something, it didn't go as planned or they found a need to augment some bit of technology and then that presented the
  • actual, you know, reason for being as a company, right? So that's really interesting and I have to ask they usually ask us at the start loft labs is there is something behind the name at all or how did that come about? I think it was just a good sounding name. I think we were in a loft building loft at the time.
  • So you know letter but no no deeper meaning for for the individual letters. We have to come up with something at some point. Yeah. Like what does loft stand for? L O like something like that. No literally you're in a loft building building this product which is awesome. It's like, you know, Apple could have been garage if if if you're talking about the same sort of thing because Apple was Apple was famously built in the garage.
  • All right, so we got we got a little bit of time left. Five minutes. Where do you see this going? So, you know, Kubernetes, like I said at the outset, people have stopped talking about it from that hype curve perspective. There's a little bit more of an of acceptance that it's just in in the background murmuring along.
  • So, where do you see this taking Kubernetes? I think you're solving a real problem here of that multi-tenency which was a huge issue for me huge issue for a lot of people. So where is Loft Labs going to be going in the next couple of years? Where do you see Kubernetes settling um in in in itself actually in a couple of years? Yeah, it's wild thinking back and it's you know it was just Kubernetes 10y year anniversary right maybe six months ago or so. Yeah.
  • And and it's it's fascinating uh how when you go to CubeCon, there's still a lot of energy there, but you're right, it is not as much of a hype topic anymore. It it is the business as usual kind of thing. And there's innovation still in in in the space, but I feel like they're more on top of Kubernetes and maybe with VNO below Kubernetes, but around Kubernetes rather than in Kubernetes itself because Kubernetes is a fairly stable system.
  • And despite the fact that I'm saying this, upgrades are still hard for people in Oh, yeah. every day that that they're still struggling with the three annual upgrades of Kubernetes version but we don't have I think these um I think Kubernetes has a relatively stable API at this point and there's a stable set of first layer products you know think of things like service mesh and you know for example or um you know admission control like there's solid frameworks around these problems in Kubernetes But there are still higher level
  • problems especially when it comes to you know GPUs and Kubernetes more modern workloads like job based workloads and Kubernetes rather than long running workloads. I think there's still a lot of innovation happening in that space and for us we're obviously much closer to the infrastructure layer than the application layer with with naturally what we're doing.
  • Um but we see a lot of this happening and we see um a lot of folks building on top of our technology which is really exciting. I think there's all of these providers whether it's like a spectrocloud or ref or platform 9 these traditional cluster managers they've all integrated vcluster into their products building on top of us right for certain big fan of platform 9 um I was there almost at the start of that helping helping helping sesh do his thing he's been on the show as well so yeah massive fan yeah they're still they're still
  • going on yeah it's really cool yeah it's it's awesome to see whenever I spot it in in any kind of docs, right, and I see them run examples and and tell their their audience effectively to, you know, work with vcluster or I see folks in the rancher community demanding opening issues in the rancher repo.
  • Hey, can you build a vCluster integration as well because like Spectrocloud has this too and we would love to manage virtual clusters just like we manage namespaces and regular clusters with rancher. It's something we announced at coupon as well. I don't think Rancher has listened to their community over the past years and built that integration.
  • So, we just went ahead and built it. We open sourced it last week. And I think we're trying to bring virtual clusters into more spaces whether it is through collaborations or us just going ahead and building a piece of open source tech that you know the respective uh communities can run with because they plug in this extension into their existing platform.
  • But we think more virtual clusters in more places that's that's a net positive for the world. So we're trying to make that happen. Good stuff. So you know what what what's shortterm? We've got about a minute left. So you know give a little bit more about what you're doing in the short term um in a minute and then we'll wrap up.
  • Yeah, I think in the short term there there's a couple of really exciting topics. uh one thing is actually having dedicated nodes mounted into the virtual cluster which sounds ironic because we went away from dedicated nodes right yeah but for very specific cases people have asked for that and what I'm particularly very excited about is once we solve that challenge because we don't want to actually have a completely separate scheduleuler in the virtual cluster so we have to figure out how how we're doing this um but what I'm excited
  • about is the end goal of saying okay you can also do it in hybrid You can have the regular multi-tenant way and these shared nodes plus I can have this one particular dedicated node. Just think of like you're a GPU cloud provider and you're saying oh we bought a dedicated one for specific important workloads but then we have a lot of testing and deaf workloads right and things where we're experimenting that can run on the cheaper shared infrastructure.
  • That's really fascinating to me. That's what we're working towards in the in the near term. Brilliant. Well, hey Lucas, I mean this is great. you've solved a problem that I thought about Kubernetes like when I first was playing with it. That is fantastic. Love what you guys are doing. Um it's been a really great conversation.
  • I'm sure we're going to hear more about Loft Labs and the technology that you're bringing and and kind of changing the industry a little bit as well. So kudos to you. Thanks for being on episode 98 of Great Things with Great Tech