 My name is Thierry Carras. I've been involved in a number of open-source projects in the last 20 years, starting with Gen2 Linux, then Wuntu Server, and then for the last 12 years working on OpenStack, first as its release manager, and then I got hired by the OpenStack Foundation, and I'm now the general manager at the OpenInfra Foundation, which is the new name for the OpenStack Foundation. I'm also a vice-chair for the Open Source Initiative, the body that oversees the Open Source definition, and today I wanted to talk to you about OpenStack and its relevance today, because when I usually introduce myself like that, like people are, or are you still working on OpenStack? I mean, OpenStack is dead. I mean, like it's so yesterday, and like just yesterday I overheard the discussion in some other talk where the people were talking about OpenStack in the past tense, you know, like, well, it was killed by corporate interest and all of those things. So a lot of people think OpenStack is dead, so I guess it's a fair question, right? Like, it's been around for 12 years, nobody mentions it's in the press anymore, so I don't know, maybe it's dead. And so looking into that question, maybe, I mean, it could be dead because it has no users, right? If an Open Source project has no users, it can definitely be considered dead. But the evidence points otherwise, because objectively, it has a massive user footprint. We run a user survey for this year, and the results are just in, and the survey revealed more than 40 million CPU cores of computing power being driven by OpenStack. That's a massive usage footprint. And you can see it in, like, areas where you might already know that OpenStack is strong like telecoms, or web and entertainment like Salesforce, Adobe, Blizzard, Ubisoft, Bloomberg, Workday, all of those companies, you know. But we are seeing a lot of growth in the public cloud sector, and I'll go get back to that, especially in Europe, where there are companies like OVH, Cloud, Deutsche Telekom, Binaural, or Chlera, all of those European-based public clouds. We also see a lot of growth in the academic and research area with a lot of supercomputer or academic collaborations being built on OpenStack as a bad one. So it's a massive footprint. It's a growing footprint, because the last user survey from the previous year in 2021 revealed 25 million CPU cores. So you can see that it went from one year from 25 CPU cores to more than 40 million cores of CPU being driven by OpenStack. So it has objectively a massive usage footprint. It's also growing in the super large deployments. The survey also reveals a lot of instances where OpenStack is driving more than one million CPU cores in a single organization. So Line, for example, which is a messaging app in Japan, Walmart Labs, which has been using OpenStack for a very long time. China Mobile and China Unicom, which are like the big two big Chinese mobile phone companies, mobile network companies. Workday and Yahoo, all those have reported more than one million CPU cores of deployments on OpenStack. So it's definitely not dead from a usage perspective. But maybe it's more like a zombie. It's dead, but it's still working. But inside, it's no longer live. There is no development going on. There is no open source activity. So in order to compare, I guess, if you look at a very active, very popular project that you might have heard of called Kubernetes, if you look at last year and the CNCF DevStats, it reported more than 33,000 GitHub pull requests being merged. Means like a human reviewing a change, accepting it, and being merged into the code base. And that's a massive, it's a massive level of activity, like only a couple of, a handful of open source projects currently active reach that level of activity. But if you look at OpenStack, it actually had in 2011, around the same number of changes being merged, like a change being reviewed by a human in Garrett, and then approved for merging in one of the OpenStack sub-projects. So clearly, if being dead means being as active as Kubernetes, I mean, I'll take it. But it got me thinking, right? Why are so many people saying OpenStack is dead? Is that that they're like, evil people that want me to feel bad? Is it like just that they are misinformed? Or is there like something else, right? And to answer that question, you have to take a step back and look a bit more at OpenStack history. So we need to go back to year 2010. The top song back then was Love the Way You Lie by Eminem and Rihanna. So that doesn't sound like too long ago, but that's actually a very long time from an open source history perspective. So for example, when OpenStack was created, the term open source was only 12 years old. And we moved 12 years since. So it's basically the middle ages of open source. Firefox was eight years old. Ubuntu was six years old. Git was five years old. Android was only two and a half years old. So all of those things that we take for granted today were very young back then. And if you look at those middle ages of open source, today open source is ubiquitous, it's everywhere, but it wasn't the case back then. When OpenStack was created, really open source was not present in the infrastructure space. Like if you look at who was providing infrastructure for like a deployment of applications, VMware was basically owning all the private sector. It was clearly 99% of the enterprise deployments were running on VMware. And if you look at the public usage, well, it basically was Amazon web services. There was just no other gaming term. There weren't as large as they are today for sure, but there were clearly no other alternative to public clouds. And that's when OpenStack was created. It was famously created by like a collaboration between NASA, bringing the Nova component, which is the compute VMs component as a service, equivalent of EC2 basically, and Swift, which was built within Rockspace for providing object storage. So equivalent to Amazon S3 technology. And the two projects combined and formed OpenStack to deliver a comprehensive cloud infrastructure solution that would be open source and openly developed. And so that's when things started to explode. It was like nicknamed the Linux of the data center. It was like the technology that would end all technologies. We would forever use OpenStack and everyone got very excited. Startups everywhere, money flowing, you know, lots of people trying to make profit out of it. Great. And that's when like OpenStack development peaked around 100,000 changes per year, which is like three times bigger than Kubernetes is today, which is like a bit mind blowing in terms of levels of development activity. And the reason behind that was that everyone wanted to be a part of it. Everyone wanted their use case being served. Everyone wanted their product to work with it. That drove a lot of, I would say, extra development activity that wasn't really required for the core use case. And so with that came scope creep, OpenStack expanding in areas where it was less strong like application orchestration and all of those things that like moved it away from its core function of providing those infrastructure primitives for others to build on. And that might remind you of other projects that are going in the same direction with the level of attention that they are getting, but I won't name names. And finally, beyond all of that, there was the question of who is the user in the end? Like is the user the people that would consume the OpenStack APIs, like the people that would use directly an API call to spin up a VM? Was it the person that would deploy OpenStack on top of a number of bare metal servers to provide that infrastructure? Is it like something else? There was a lot of confusion when we talked about who is the user. And that's when we got a welcome clarification, thanks to the rise of Kubernetes. And in order to explain that, I'll take another step back and look at the way applications are provided to users. So if you look at 20 years ago, the way you would provide applications to users is by procuring some physical hardware and then deploying an operating system, the dependencies, your application on top of that. And so that was clearly a lot of work. But we started to add more and more layers on top of that. The first layer that was added is clearly hardware virtualization, basically abstracting the server your application is running on from the physical server that runs it. And then we added cloud APIs like OpenStack to provide a programmable interface towards those virtualized resources. And then we added application deployment APIs, which is basically what Kubernetes provides. So a set of APIs and primitives to deploy complex applications on top of that programmable infrastructure. And really Kubernetes created this layer between people whose role it is to provide infrastructure for others to depend on. So people that would provision a Kubernetes cluster on top of programmable infrastructure and the people that deploy and develop applications that would just consume those clusters and deploy their applications on top of that. That clarification wasn't there when we started in 2010. And so this layer really clarified that for those infrastructure consumers, it's someone else's job to provide infrastructure. It's no longer their job to care about it. So developers, deployers, for all of those people, infrastructure is a given. It's really someone else's job. And so they no longer talk to OpenStack directly. OpenStack is relevant to them. OpenStack is invisible to them. And so in a way, OpenStack is dead to them. But it's not dead. It actually just grew up and it found its purpose in life. It found its user basically. We know that now that the user of OpenStack is the infrastructure provider, the person that will deploy it, choose to use it, deploy it, and then provide those resources for everyone else to build their applications. And around OpenStack, we've seen this rise of this whole class of actors separate from the traditional developers whose role it is to provide infrastructure within organizations, in public clouds. All of those people work invisibly to a lot of others to provide that infrastructure where everyone else is depending on. And recognizing that this is a separate role and like to propose more open source solutions around that space to that specific population. That's why the OpenStack Foundation transitioned from being centered purely on OpenStack to be more inclusive of other open source projects. That might be relevant to the same population. And so that's why we support more than just OpenStack today. I won't go through all of them. That would take too much time. But I want to mention Kata containers, for example, for secure and lightweight virtualized containers. If you want to add security to your Kubernetes mix, that's probably a good solution. And Starling X, which is the distribution of OpenStack and Kubernetes tailored for edge and industrial use cases. So all of those things make sense for using more open source in the infrastructure space. But back to our topic, back to OpenStack. If OpenStack is not dead, what makes it relevant? Like what makes it interesting for the next decade? It's good that it's successful. It's good that it's 12 years old and still around. But why would it be still relevant 10 years from now? Like shouldn't all of us just use hyperscalers and run our workloads there and don't care about infrastructure? That would make a lot of jobs easier, I guess. Well, so if you look at what happened over the past years, we really saw the emergence of a new open source infrastructure standard. And that's the rise of what we call the Loki Stack. So combining Linux at the bottom, OpenStack on top of that to drive hardware, Kubernetes clusters to be deployed on top of that programmable infrastructure and giving developers access to that infrastructure to run cloud daily applications on. And we've seen that pattern everywhere. This end-to-end open source solutions for providing infrastructure is really interesting for a lot of people, especially that could be in the public cloud space where everyone can request those resources using their credit card or in the private cloud space where it's limited to the use of one given organization. And if we look deeper into the why, we built this value proposition a few years ago saying that OpenStack is relevant because of cost, compliance, and capabilities. And that was probably in 2017 that that came out. And it's actually more relevant than ever. And I'll dive into specific cases. So if you look at cost, for example, there was this study by Andrews and Oritz around the cost of cloud and basically concluded that rapatriating workloads from the public cloud to the private cloud resulted in one third to one half of the cost. And that's because of the giganormous margins that the hyperscalers are taking on the public cloud we are all consuming. So it might be like, it's not for everyone, it's not for all workloads, but clearly there is this sense that a hybrid infrastructure makes sense because if you have a usage that is obviously moving up and down in time, you should probably invest in private infrastructure to serve the baseline of the workload and leverage the elasticity of the public infrastructure to handle the SPACs. If you just use public infrastructure for all of it, you will just bleed all your money to Amazon and Microsoft and Google, make them very happy, but make your CFO not very happy. And so combining those two is really critical and a VC company like Andrews and Oritz seeing it is kind of interesting. They're basically advising their startups to think about rapatriating really quickly in order to keep their money. So the second aspect was compliance. And compliance is today, especially in Europe, a topic that is extremely important. We are seeing the rise of the digital sovereignty concerns really driving a lot of new usage for OpenStack. When I said we moved from 25 million to 40 million CPU cores of computing power using OpenStack over a year, a lot of the new users we are seeing is driven by those digital sovereignty concerns. So it drives a lot of new public clouds, especially in Europe or Southeast Asia, countries that don't want to depend on China or the U.S. for hosting their workloads, especially in today's troubled times. That drives a lot of new usage for OpenStack. Because how would you start from scratch? Well, you should probably use what's already developed and that's a lot simpler. We're also seeing it in a lot of research clouds. So on the slide earlier, there was the European Weather Association, the ECMWF, which is an associate member of the OpenInfra Foundation now. And they are running all the models for predicting the weather in Europe using an OpenStack cloud that wasn't there five years ago. They used to run on the public cloud, they moved. And that's also so that they can maintain their independence, maintain their EU funding, which is coming with more and more strings attached. We've seen the announcement for LF Europe this morning. All of those things really point to a need for workloads that would run under local jurisdictions and not be potentially vulnerable to things like the Cloud Act in the U.S., where they can just shut down your workload or access the data into it. The third aspect was capabilities. And what we've seen is a lot of innovation in use cases where OpenStack is being deployed. You could think that cloud was invented in 2007 and now it's very clear cut. There isn't just one way to do it and Amazon does it. And so why would you need specific things that would require specific providers? Well, I'll go into a few examples. So one code is... Oh, that doesn't show well, but you get the guess. So it's basically a public cloud company that built a server data center in Guam in the middle of the Pacific Ocean. And their use case is very special. They want to have equal latency between their data centers and Japan and China and the U.S. and Australia so that they can host game servers and host eSports competition. Not the use case that Amazon would build a data center around, sure, but a small company that can deploy OpenStack, why not? And so this is a very specific use case, a very specific innovation that they are wanting to... Well, innovation. They want to drive through this and OpenStack enabled it. If OpenStack wasn't there, there would probably not be this company around because they probably wouldn't develop it from scratch. Another example closer to Europe, closer to France, Guam based, is Exion. Exion is a startup within EDF Group, which is the national world top electricity company in France. And they're running nuclear plants. The EDF is running nuclear plants. To run a nuclear plant, you need to have super computers modeling how the system works. And they rotate those super computers really fast. And so they were basically scrapping probably hundreds of millions worth of super computers every three years in EDF. And there was this group within EDF that said, well, we should probably reuse those. We should probably recycle those. Those are very powerful machines that can run a lot of workloads close to one another, which are like usual use cases for HPC. So how can we build something? So they formed a startup within the group and they are recycling those old super computers, installing OpenStack on top of it and offering a public cloud that does HPC as a service and also blockchain as a service. And so it's clearly, again, a use case that is enabled by OpenStack. A very corner case where they have this need, they have those machines rather than try to sell them or scrub them, which was what they were doing, they just recycle them in an OpenStack cloud and now have a very, very profitable offering. Third use case is LiveCloud. LiveCloud is based in Amsterdam, so also in Europe. And it's a public cloud that is distributed over the city of Amsterdam. They have the compute nodes in apartment buildings eating the water, using the electricity that is used for cooling, like all the heat that is generated by cooling the servers. They set up a system where they are cooling the machines with the system that eats the water. And that means they actually have poor efficiency that is unpowered because more than 50% at least, but in most of their models, 80% of the electricity that they are using for running the servers is actually used for eating the water in the end. So you end up with a very specific use case, I guess. It's all because Amsterdam is holding tin black fiber. They can actually have the latency to make it work. So it's not applicable everywhere, obviously. But again, it's an innovation that is extremely green compared to traditional data centers and that is made possible by the fact that OpenStack exists. And so in conclusion, I would say OpenStack is not dead. It has a massive user footprint. It's growing year over year. It's still one of the most active open source projects in the world in the same footing as Kubernetes. But it might be dead to you, though. It might be invisible to you. It's someone else's job to care about providing infrastructure. And that's fine. It will not replace AWS. I mean, when we started, we said, well, we'll just displace AWS thanks to a federation of OpenStack-powered public clouds. Now we're just seeing that it's still there, and Microsoft is big, and Google is big, and whatever, Tencent is big. So it will clearly not replace hyperscalers. It will not replace every technology going forward, which was kind of part of the craze at the start of it. But it is a necessary component in the infrastructure-providing landscape. It's a tool for enabling hybrid usage, which we've seen is very cost efficient. It's a tool for serving use cases that the hyperscalers will not serve, like we've seen with those examples, or for workloads that cannot be served from US-based or China-based structures because of digital sovereignty concerns. So I would say that for cost, for compliance, for capabilities reason, OpenStack is here to stay. Thank you. And we have plenty of time for questions. If you are not hungry enough. Is it on? Yes. Okay, so I have a provocative question. So I like your division between the user likes Kubernetes and OpenStack is for the infrastructure. But what you actually are missing, I think, is that OpenStack itself is also an application which can run on Kubernetes. And actually many in our company, for example, we actually use Kubernetes on bare metal to actually host OpenStack on top and to provide that infrastructure API. So we kind of have this sandwich problem. And Kubernetes, essentially provocative here, is also extending into the infrastructure model with a nice clean extension API. And you can get container native virtualization and bare metal, there's metal three, and what do you say there? Yeah, I think Kubernetes is definitely extending in more and more of providing those other primitives that like OpenStack is currently providing. I think it's a good thing. It proves that it's needed. I would say that ultimately it will suffer from the same scope creep that OpenStack suffered because if you just lose your core function, the attention to the core functions, it kind of becomes difficult to maintain the core of the system. We've gone through it. We have a lot of discussions with the Community Steering Committee about that, like the life cycle of a project and how difficult it is for them today to maintain that core because everyone is pulling at the edge to serve their use case, have a new startup about a thing. The layer we have had clear separation with, I would say, the specialty in every layer. And so you're right. We're seeing Kubernetes more and more extending in that direction. I'm not sure it's a good thing for them, but I'm welcoming more open source usage in infrastructure, whatever the solution is. The mission of the foundation is really to enable access to infrastructure technology to all. It doesn't have to be through OpenStack. It can be through other projects. We're all happy. So it's fine. Yeah, thank you. Nice answer. Things to ponder about. Hi. This is also a provocative question. Yeah, Daniel Overtalk is asking for provocative questions. So is there a lesson to learn here? Many people got very excited by OpenStack. Many were contributing, made a great impact and was sort of an essential step in the development. But was the outcome just that a couple of companies could make business out of this? So yeah, OpenStack has like a striving ecosystem of companies. I would say that there was a moment where everyone was on it. Clearly, and some people didn't have a business model around it. They just came to OpenStack because they were attracted by the light or the gold, I don't know. So clearly, there were some companies that were not ready or did not really have a business model. We've seen that happening in other communities as well. At some point, people want to be part of that thing that seems exciting and generating a lot of revenue. But in the end, you have, there is just no space for 10 red ads in OpenStack. So the market of companies that actually build a distribution on top of OpenStack was reduced to 2, 3, 4 companies. And yet, yes, the others just pulled off of the game because there weren't any profit to be made. So I would say, yes, there is a lesson that you should be aware that the hype is driving a lot of that attention and it's brittle. It will not last forever. We're already seeing signs for communities of like, it's found its place and maybe trying to overextend it is maybe not the right approach. It's still like, there's like about a five-year lag between OpenStack and Kubernetes in terms of lifecycle. But I would advise the next I mean, we would both, Kubernetes and us, advise the next big thing that will attract whatever hype, I don't know, AI, blockchain, whatever, to be very careful about not building too many systems on top of that gigantic scale that they will reach. One of the big issues we had with OpenStack is that we built system to sustain 100,000 changes per year of activity. And that's very heavy or complex systems. And when you need to scale down because you're back to a more normal level, it actually is difficult to unravel all the systems that you built to care about all those things. And that's where like manpower issues start to pile up, all your rules don't really work anymore. And it's difficult for people to just accept that and unravel everything they built, all the complexity that they built. So there are like definitely lessons there. And I mean, when I joined OpenStack in 2010, I was not expecting it to be nearly as big as it ended up being. I did not jump into it with the, at the moment where everyone was jumping into it was there at the beginning. And it's been an interesting ride, I would say. It's a really good experience, but there are lots of lessons to be learned. That's why we're directly engaging with like other projects to share what happened. And so that the next project doesn't make the same assumptions that we did. But the conclusion of that is there is a risk to be overfed. And you have to sort of push away maybe some people with too high ambition or but an open source project on the other hand is sort of it's done by the contributors and anyone who wants to contribute and sort of doing relevant stuff should be allowed in. It's very difficult to say no. And that's why the scope creep is happening. Like it's very difficult to prevent someone wanting to join your community and do things within it. Say, well, no, this is off limits. It's very difficult to start a project from day one with that extremely well defined scope and say, no, you're not part of our party that just goes somewhere else because this is just about this. We're seeing it with like some projects are very, very clear boundaries. And it's good, but it never was the platform the size of OpenStack or Humanities. With a platform it's much more difficult to say no. Yeah, thanks. All right. Well, we'll go for lunch now. Thank you all.