 Live from Miami Beach, Florida, extracting the signal from the noise, it's theCUBE, covering .NEXT conference, brought to you by Nutanix. Now your host, Dave Vellante and Stu Miniman. Welcome back to Miami everybody. This is theCUBE's coverage of .NEXT, Nutanix's first user conference. Alex Povey is here, he's the CEO of CoreOS. You guys are in the heart of things. Distributed web apps, Kubernetes, containers, all the cool stuff, thanks for coming on theCUBE. We like to call the Core. The Core, it is the new Core, right? Yeah. So, great, what do you think of this conference? It's a great conference, man, I'm here in Miami Beach, beautiful outside and quite a show, lots of- Steamy. Yeah, steamy. You know, there's a lot of folks interested in learning about how to use this new technology and so it's a great customer event for us. So you guys are kind of hand in glove with this infrastructure play. Maybe start by talking about CoreOS and where the fit is with this event. Sure, so yesterday, Kelsey Hightower, who's the product manager for us, gave a demo of Tectonic, which is our container platform built on Docker and Kubernetes and all these different components. And so, the Nutanix platform that they've been talking about quite a bit, also now supports Tectonic as a commercial partner on the container platform for them. So it was a nice chance to show off some of the new tools that we've been building and then also partner with the Nutanix guys to talk about how their customers can use these types of deployments as well. So Stu had an interesting segment earlier with Alan Cohen and JR from Cumulus and I think one of the folks on your discussion put forth, what if the future stack is not this hierarchical beast? So, what's the future stack look like? Distributed functional- Easy entries and exits into those functions. I describe it. I think what's happening is we've seen the model where we all end up because most, if you put a group of architects in a room together and get to really focus on building the right solution, they generally pop out with similar patterns and architectures and what we're seeing happen is a lot of emulation from the hyperscale guys right now, the stuff that Google and Facebook and the hyperscale platform providers sort of taking those models and applying them to everyone else that hasn't taken the time to build the teams to like build out these platforms internally in a general purpose way. And so we're taking those methodologies and containers I think really unlock the discussion around it but really it's containers about packaging and running a process and that's great but the real interesting thing that's happening is things like Kubernetes is like how do we run our infrastructure in this new way? Which is just different than the previous way of running infrastructure and we're taking the hyperscale kind of methodologies and making them the best parts applicable to enterprises and smaller companies and so on that previously didn't have the resources to invest in building that type of sophisticated infrastructure. And so what we see with companies that are maybe they're running OpenStack, maybe they're running Bare Metal, maybe they're on cloud and that's all great but you still need an operational model for how you actually run the infrastructure. It's like the up to the layer of running Red Hat plus Puppet on top. It's not just about the compute, it's about the actual operational model and I think that one of the more exciting things is that you can kind of bring together all of those different environments with this way of running infrastructure. So give us the 101 on that. So you got Linux and it's going to give you the container and you're going to put contents in there, you're going to put processes in there. Where do you guys fit and how does that evolve? Right, so just stepping back. Why are we doing this at all? So the reason we got started in this whole space is we've approached it from the position of security. We think that there is an opportunity to fundamentally improve the security of the internet. Now our approach to that is around serviceability, making it really easy to patch and update your infrastructure and so the model that allows you to do that is this distributed systems container approach. Now there's a bunch of other benefits about running infrastructure that way around high availability, utilization, bunch of other things but we got into it thinking about with a really security lens up front and saying this way of running infrastructure will allow us to build a much more secure piece of infrastructure. And so that's sort of how we got in the middle of all this and we started building the individual components required to enable this new world. So that was Core West Linux to start alongside SED, our distributed data store, brought in Flannel, our overlay networking technology, we brought in Fleet which is early versions of this container orchestration stuff and then around us we've seen the Docker ecosystem boom, we've seen Kubernetes boom in its own right and a lot of that is the startup luck piece in that the market is sort of merging around it right as we are putting together a lot of the enabling components to make it happen. Yeah, so Alex, one of the things coming out of this show and something we've been watching closely is for the last 10 years, that virtualization layer was almost at the center of the universe, VMware leading that charge, Microsoft's come on there, now you've got cloud, you've got containers and the center of the world is really getting more towards the application. So can you talk a little bit about what you see at this show and overall as to how we're moving along that journey and I mean there's some big air begups and challenges to get from kind of the traditional way to the new way and how are we going to get past that. Right, so I think the key change that happened with containers in particular is that focus from operating a server to operating an application and it's a subtle change but when you started thinking about the job of an IT and operations professional of operating applications instead of operating servers the way you can manage it changes fundamentally and so that's really the thing that's happening right now. Servers aren't going away, virtualization isn't going away, you still need all the stuff to run applications just looking through a different lens about how we do it and again you get these benefits around security and utilization, all these things if you're focused on the application first instead of the server first. That's what we think the big change is all about with this. Yeah, I wonder if you can talk a little bit about when you started your business. For a while we've been talking about open source is going to eat software and it's changing everything but how we actually monetize that has changed a lot. I had a good conversation with the technical folks here at Nutanix and said, well Acropolis is based off of KVM and how do they relate? Of course they make changes, they give them back and it's there but they're not coming here saying we're an open source company and we're going to sell you an open source hypervisor it's their platform, they're sensible, they've got an ecosystem so what's your viewpoint on the role of open source and how does the company actually make money there? Sure, so first every product built today, open source, you're not, it's built on open source. Okay, if you're GitHub, is GitHub an open source application? No, but is there a lot of open source there? Yes, if you're Amazon Web Services, is that open source? No, but is that powered by open source, definitely. So when we talk about open source business models it gets confusing because it's like how do you sell software, that's free. Well there's only really three things you can do. One, you can sell software, you have commercial software that's built on open source but in and of itself it's proprietary, so that's selling it. You can provide support services on the open source side, that's another option. Or you can provide hosted services, so more hosting services version, software as a service version of the open source. See these database companies that will run database before you and there's a bunch of other things like that. Or four, you can talk about it on there. Or you can talk about it, yeah exactly. In the sense we are a software company and we have an advantage on building software. That's what our company has an advantage of and the three ways we can make money off of that are by again, support and services which is the most basic one but hosted services or commercial software. And we have all those pieces that we do. We don't think we need to be a one-track pony on that front and we're not afraid to build commercial software where it's appropriate. We do treat open source software as a way to just move the industry forward. There's just basic internet plumbing that needs to exist in order to move everybody to this new way of running infrastructure. If it was all proprietary we wouldn't see why it's right spread adoption. Meaning you're putting a lot back. Yeah, I mean our core business but our core thing we're known for is all the open source software that we put out. Like, etcd is a great example. It's a distributed key value source and underpinning of building a distributed system. And it's, we have no intentions of directly commercializing nothing at all. There's never going to be a community edition versus an enterprise edition of etcd. It's just basic internet plumbing that we need to exist in the world in order to help companies build distributed systems and we want companies to build distributed systems. And in the case of etcd it's happening. Like Kubernetes is based on etcd. So in a way we help enable that product. Now they did a ton of hard work there, don't get me wrong. And a lot of companies did a lot of hard work there but we contributed in our own little way of making that effort more reasonable to build. What makes etcd more secure? Can you talk about that? I mean, you know, fundamentally I'm more familiar with the cumulo and granular levels of control but can you take us through that? So etcd in and of itself is more on its unique value is not security. It has some security features but the main thing that it solves really well is the hard part of building a distributed system, a consensus algorithm, which is a PhD thesis from Stanford type thing. And it's one of these pieces that's really, really hard to get right. And once we get it right, let's just like have one that we all share. Let's not go build our own independent versions of it because it's really difficult to actually get it right. And so etcd solves this consensus algorithm problem and exposes it in a way that makes it easy for other developers to consume. Yeah, I tell you Alex, the first time I heard about that it was like, wait, this is doing one of the top level things that we've been talking about for etcdn for the last bunch of years but this does it, it is there and it's one version. So with Flannel is our lightweight etcdn technology that we built to help containers get overlay networking and it's as raw and simple as it gets, it's backed by etcd for all the routes and how the containers connect to each other. So you have a distributed, highly available data store for the really important parts of the data and then all the overlay is just in kernel like using the technologies that already exist today. So it's this very lightweight shim that gives you the data requirements in this extremely highly available way but then reuses as much as the open source size possible, all the things that companies have been investing in really, really heavily to make the overlays fast. And it's just like these two components paired together, you get a super lightweight and a performant solution and easy to set up and so on. I'm wondering if you can just give us your viewpoint on kind of the container marketplace. I mean, I still talk to people that say, containers and like Docker, you mean the pants type stuff so don't know what I mean. We've been living it for the last year to two years so those of us that have watched it have changed but security, networking have been challenges. Most applications are under their stateless today. Where are we in this journey and how fast are we going to really get to containers everywhere. So first off, if you look closely, things like rail six, which a lot of enterprises are still running, already are running everything in containers. They just don't know it because they're not taking advantage of the features but the applications are already running inside of containers using Linux C groups and namespaces and this stuff. They just don't actually have it all turned on and enabled in the way that like a Docker or a Rocket enables it for you in a full more isolated mode. So people already using it don't even know it because they're just existing Linux kernel features that have been out there for a while now. Now, the big thing that's happening right now is this new way of running infrastructure. We call it Google's infrastructure for everyone else. If you think of a container, yeah that's a package, that's a way of running a process, that's great but this platform, this way of running infrastructure in a distributed way I think is where kind of all the excitement is at. Even Docker's building things, Mesa's building things, Google's building things in this space so that's kind of the big overall shift of the Delta but we're still early. I mean this really reminds me of about 2006, 2007 of AWS. Do you remember when we were all doing, there was the AWS event and then there was the cloud events and we all had our conferences for cloud and it was exciting. You wrote the paper above the cloud. Exactly. Some people talk about cloud, some people were talking about EC2 and it was like. Larry Ellison was saying, that's cloud. Exactly. And I think while this containers have spread very widely in the industry, very quickly, I think we're on a normal technology adoption curve here. People are putting it in production but it's still something that isn't, we're not yet at the point where enterprises are wholesale, moving all their infrastructure over to these new platforms. One, because the platforms aren't done yet and two, again, normal technology adoption curve. We're just, it will happen over time but we're in the early days. So maybe, I mean, what milestone should we be looking for, where's the big activity in the various, you've got so many projects you guys are working on and the solutions you're putting together like Tectonic. So our goals right now as a company are really bringing together all the components. We think all the necessary pieces are out there now from CoroS Linux up to Kubernetes and everything in between, Head CD, Fleet, all these pieces. We think all the components are there. Now we need them all to stabilize and get well integrated into a cohesive thing that's actually really easy to use and can actually be put into production in these environments in a rigorous production way. This is not for people that are okay, cutting corners, downloading random stuff off the internet. These are for real enterprises that need like production ready, kind of rigorous requirements of all security and availability. So you're saying the future is here and it's distributed but it's not all pulled together yet. Yeah, the future is here. Yeah, it's happening in terms of like every major IT vendor right now is starting to figure out how do we teach our customers to run infrastructure in this new way. In this way we're going to, we call it GIFI, Google's infrastructure where everyone has this like platform, containers, distributed systems, commodity, machines. Commodity today doesn't just mean cheap hardware, it means virtual machines on AWS. You know, it's running your infrastructure in that way and it's going to happen over time here and all the pieces are emerging right now. And you're right, 100% of the companies out there are trying to make this happen if they're not, they're toast, so they are trying to make it happen. What do you call it, GIFI? Yeah, I get tired of saying Google's infrastructure everywhere. So GIFI, so is that Google's infrastructure circa 2015 or is it Google's infrastructure 2011 that people are trying to bring to the enterprise? I was wondering, what are they working on now that we're going to be talking about in five years? I think there's a piece of that. I think for a while the competitive advantage was actually on the infrastructure and that's loosened up a little bit now. And so the research papers that these companies put out are like getting notes from 10 years in the future about what's coming, you know? But there's new stuff, I think data is one of the big problems that still isn't solved as much as there's talk around big data, things like DynamoDB or the data stores that live in these cloud providers. H-based didn't solve that problem. Yeah, they're distributed, but they're not highly available in the same way. They can't be on scale right throughput in the same way. They're not gonna be secure as you are. And so those sorts of technologies I think we'll see, but the way, the model of running infrastructure kind of drives all of that. Those types of data stores make sense when you have these scale up, scale down, more ephemeral type workloads where you need this globally wrapped data store that is just always there and can scale right throughput with capacity as well. And so that's some of the stuff I think is coming, but another thing I think is we're, I think we're going to be able to surpass at least in a quality level what these bigger companies have done because it was a secondary focus for them. They built this as a necessity, not because they were trying to ship it as a product in it of themselves. It was a side effect of their team trying to deploy software. In our world, our product is this way of running infrastructure. So we can focus on just doing it all right. Yeah, your why is to change the way things are done. Their why was if we don't do this, we're screwed. Right. So that's a different mindset. So, but you're saying, Alex, that essentially that model is fossilized, that basic scale out model is fossilized and it will evolve over time and will the enterprise and the web scale guys sort of evolve together in your opinion or will the web scale guys still have this five year lead? So, the web, the hyper scale guys are going to stay ahead on their own front and they've already made their bets. Like there's no way that Google's going to be running tectonic anytime near, because they've just built so much of their own infrastructure and you know the way it works. But at the same time, I think these enterprises, once they make that shift, they'll be in a position to much more rapidly make the next change because your infrastructure becomes something you're comfortable changing and upgrading and making better. And so that once we cross that into this world of running infrastructure, I think we'll see further acceleration into new ways of adopting things because we no longer have to treat our infrastructure as you know, don't touch it because something's going to break. Like we actually solve the hard problems that allow us to break things without breaking the entire environment. To let service fail without taking anything down. So the collective leverage of that enterprise economy is let's call it two trillion of spend is actually greater collectively than the combined web scale. The economy, if you add up Amazon, Google, Facebook, Twitter, LinkedIn, whatever else you want to throw in there. The enterprise, even though it's diffuse, right? You see that coming together in a way that you can achieve that flywheel effect that Amazon always talks about, such that innovation can actually come from the enterprise side as opposed to just the web scale guys. Sure, I mean, so I think an interesting question is. Kind of a loaded question I know. Well, when companies run, so if this way of running infrastructure gave a Google an advantage that allowed Google to exist in the first place because it was technically difficult, nobody else could do it. What happens when that becomes trivial? Just in the world. It's just like interesting thing to wonder if everybody can build these types of hyperscale applications trivially. Like, what do we just unlock? I mean, there's some sort of social mission there, right? Like, let's just put that hornet's desk and just see what happens in the world, you know? And I think it'll be very, very interesting. Now, for enterprises, look, enterprises are notoriously slow to adopt new technologies. It's going to take a while and there has to be super strong business value. And I think that's all there and it'll happen, but it's going to take some time. Yeah, so Alex, you know, we've been talking to a lot of the practitioners here and their roles changing a lot. They're now like an infrastructure architect, but I wonder, you know, at CoroS Fest that you had recently. It's a very different beast that you had there. I mean, there's some people here we've talked to, you know, DevOps focused and kind of span that gap, but what do you see in the roles of, you know, the practitioners that you run across? Sure. So this kind of goes back to the operating and application versus operating a server thing. So I'm assisted minute heart. I got my career running infrastructure, you know, doing this up being on call. My first company was building monitoring and management tools. I was trying to make it so my fellow ops guys were on call anymore and I built a monitoring company. So that didn't really help. But in this, like, I believe what the software is doing is helping take the dirty work out of being assisted men. It's not eliminating the job's assistants. You still definitely need people to operate your infrastructure, but it's taking out that need when three in the morning, your server goes down and right now you get a page because you have the system monitoring all of that, you get a page and your ops guy goes and makes that back to how it should be. Well, now we have a thing that says three of these should be running and if it dies, the system just brings back the third one. You know, so it's like the way it should work. And those are fires for ops guys. Those are not things they want to be doing. No ops team wants to be up at three in the morning bringing their database back online. Yeah, absolutely. And so we are working to eliminate the fires through automation so ops teams can focus on moving their infrastructure forward. Not have to do this stuff that this, you know, patching is another one. We automatically stream or patches and updates to CoroS Linux. And it's like, again, if your ops team wants to focus on delivering patches, you can still do that. We think we can do it in a painless way that you just don't have to worry about it anymore. And when there's a kernel vulnerability that has a security update, that's a fire drill for everyone. There's no one sitting around for that. So let us just take care of it for you so you don't have to worry about it at all. Yeah, we had Bob Love from a bottom line on theCUBE yesterday and he said, you know, Nutanix helps give me my weekend back. And boy, I know so many people that, you know, patch Tuesday, if we can eliminate that, you know, boy, there's the stuff that the CIS admin is saying, I'd be really happy if I don't have to do that job anymore. Let's just put out the fires. Just focus on the business. With software. And let you go back to your job, which is architecting and building out these environments that solve your business problem. The business problem is not the kernel patch. That's just a side effect of having your business problem, you know. All right. So can you give us just kind of the business update for CoroS? Were you guys, how many employees? Talk about customers and that discussion. So we're just under 40 employees. We've raised 20 million in capital so far. Our primary investors are Kleiner Perkins and now Google Ventures. They recently got involved. Sequoia's in. Yeah, Sequoia, we caught them all. Oh, what's going on? Exactly. So they have great investors, Silicon Valley's finest. And then as part of the tectonic announcement, which is, you know, this complete package for companies that want to get going on this, you know, container distributed system revolution that's happening. Well, we have Google's infrastructure for everyone else brought to you by CoroS, Google, through Kubernetes, you know, and backed by Google Ventures as well. So it's on the business side, the companies are now able to get that package and we're working on an early set of customers to help get going. Our kind of build methodology is we're going to take all the open source components. If you want to build your own tectonic, if you want to piece together everything yourself, we'll support you on the individual components too. And that's where we get into more of the traditional support business. But if you want the package, because we think we sincerely just picked the best components out there and we're going to maintain them for you and make them work really well, you can have that today as well so you don't have to worry about piecing it together. And just leave it up to the customers and decide how they want to go about it. And you're actually selling today in those three business models that you laid out earlier? Yeah, we are. So on the support and services side, we're selling on CoroS Linux, you can buy support contracts and that gives you also tools for managing your updates internally, this kind of over-the-air update model for your own infrastructure. On the hosted side, we have Quay.io, which is a container registry. We also sell that as an on-prem version of Enterprise software. So again, all three of those components just right there. And then Tectonic is a hybrid of those as well. It's a full package. Those other pieces we see as components of Tectonic, they are just, you know, if you are running the full package, you need support on CoroS Linux. You need a container registry in there in the full package as well. So Tectonic is the kind of full thing, but if you want the pieces along the way, we'll sell it to you that way as well. Awesome, what's next for you guys? We're on the world server so we can secure the internet. That's the plan. And we're just continuing to charge away at that. Next event? Where are you going to be? Next event? GM World? Oh, marketing team back home. What is the next event? Actually, go to CoroS.com. DockerCon, yeah, DockerCon's right around the corner. If you go to coros.com slash community, you'll see a list of events coming up all over the world. Yeah, it's, every week we have people all over the world giving talks about this stuff. Just curious, when all your projects on Git, what percentage is kind of internal development versus the community? You know, on the core projects, we're really a helping guide and lead those. On overall, between our, we have like 100 repositories on GitHub, I'd say we've, this is just back to the envelope. I bet we have around 2,000 contributors overall to those. And so it's a very healthy open source community. Our internal tech leads on our projects that are like CorosLinux, Rocket, EtsyD, Flannel, those are kind of the ones we are spearheading, but then we're heavily contributing to the ones we use as well, like Kubernetes, SystemD, Linux, Kernel, you know, these are all things that aren't Coros projects, but we rely on them, we contribute too heavily as well. All right, Alex, listen, thanks very much for coming to theCUBE. Congratulations on creating that feeding frenzy. Silicon Valley's finest. It says a lot about what you guys have and where you're headed. So thanks for coming to theCUBE. All right, thank you guys. Thanks for having me. All right, keep right there, everybody. We'll be back with our next guest right up to this. This is theCUBE.