 So I want to thank you for joining us this afternoon first. I'm going to be talking about OpenStack and containers on ARM in one second apparently. Uh-oh, wrong screen. Ah, there we go. Okay, great. Now we can start. Hey, Hammer. How are you? Yep. So my name's Jeff Underhill. I'm the director of OpenSource Enablement with ARM. I'm joined by Martin Stadler, who works with an organization called Lanaro. And we'll jump right in. So just a little bit on the name ARM, right? It's multiple things. And I want to make sure that we level set because we talk about ARM as a company. So I work for ARM. We were acquired by SoftBank last year. ARM is also an architecture. It ships billions of chips a year, powering a whole range of devices. So you guys are probably all interacted with multiple ARM-powered devices today. And then it's an ecosystem. We don't do what we do as a single company. We have a network of companies that we license intellectual property to. They make chips, boards. There's a whole software ecosystem. So it takes an awful lot of people to take what we develop and actually get it to market and then in your hands as end products. So some of you have probably seen a lot of this here this week. The beauty of the ARM architecture is we have very, very small, tiny cores that are embedded in intelligent devices at sensors, car braking systems, microwaves, you name it, all the way through to very high-end, out-of-order, multi-scalar processors that go into service systems in the data center. And literally anywhere in between. And what we're seeing is everybody's really looking now to run the right thing in the right place at the right time as efficiently as possible. And in order to do that, some of that intelligence has got to spread out into the network and get closer to where the data is being generated. So you can groom that information, make localized decisions. And this is something we talk about as an intelligent, flexible cloud. So running the right thing at the right time in the right place. So Martin and I gave an update at OpenStack in Barcelona last year. This is my chart. I'm not going to go through this with you, but I just wanted to, you know, what should I let these people know? Well, here's a bunch of stuff that's happened since OpenStack summit in Barcelona. These slides will be published later so you can see it. But I did want to draw attention to just a couple of these. So this is an announcement that we made with Canonical at, or just before OpenStack, Barcelona last year. We've been working with Canonical on their Linux distribution form for a number of years. Last year we added the OpenStack and Cephe storage products to that portfolio. We also work with the Suzy folks and at SuzyCon last year they announced availability of Suzy Linux Enterprise Server and their says Suzy Enterprise Storage products. And then last but by no means least, working with Red Hat as well on their Enterprise Linux. They have a developer edition of this that's been available for a while, a developer preview, you can go download that. So whichever platform you happen to be based on, the upstreams for all of these as well. So Fedora, Debian, et cetera, OpenSuzy are all available. And that gives you that kind of foundation to do your OpenStack work. So Switching Gear is a little, Linaro is a company that's essentially an area where all of the on partnership can collaborate and develop OpenSoft software and contribute to the upstream. They have multiple working groups within there. Here we've got the Home Group, the Mobile Group Networking and Enterprise Group. Martin and I primarily represent the Enterprise Group. And one of the things that we've done with Linaro too, in order to ease access to hardware and enable developers to do what they need to do, Linaro has a developer cloud that they've stood up and they're actually powering this using OpenStack. So they consume and eat their own dog food if you will and use this as an environment to consume a lot of the work they do to enable developers to have access to ARM server hardware. And with that I'm going to transition to Martin who's going to tell you more about OpenStack containers and the things that we've been working on since Barcelona. All right, great. Am I on? I'm on. All right. So I can see a couple of you in here who I know already know ARM server. How many of you have done ARM servers yet or are you just looking? A couple? All right. So Linaro Data Center suffered to find everything on ARM. That's really what we're going to be talking about. So who is Linaro? Number one, why am I talking to you? Linaro was an organization that was set up by ARM to develop open source software collaboratively. So all of the SOC vendors, the distribution vendors, everybody working together. Why? So that we're not duplicating effort. Linaro's member funded. We don't, we're revenue neutral, profit neutral. All the donations of all the members that come in we put to use into engineering, into labs, et cetera. Founded in 2010, started off with six members. Now we have 30 plus and it just keeps on growing. So Linaro in the data center, who's actually in it? So number one, we're vendor neutral and this is the Linaro Enterprise Group. All right, now I can hear myself better. Right? No, no, it's better. So we have an engineering roadmap that's defined by the steering committee members who are members of the group and we work on core open source software. So for OpenStack, you know, the group that's sitting in front of me here, really if you look at, you know, called out OpenStack, Cep, containers, Kubernetes, you know, DPDK, that's really what we're working on. But we also have to do things all the way from, you know, UEFI for ARM server that you need in OpenStack for your guest images, right? So there's work from the firmware layer through the kernel up into KVM, Q and U and then up into OpenStack and big data. That's really where we are. These are the members today that are in that group, Fujitsu, SocioNext from Japan, Aliyun, Zalibaba from China, HXT is China, ARM, as you all know, out of the UK, now also Japan. We have a couple hyperscalers. One's publicly announced themselves. That's Facebook. There's another one out there. HiSilicon, Cavium Red Hat, HPE, Qualcomm. So it's a good mix of either people that are making the chips, those that are going to actually manufacture them or those who are going to use them. So it's a nice mixer. So that's what the ARM ecosystem kind of looks like. If you look at it, you know, you've got the central area of ARM and then you have all these different companies, all these organizations that are working on open source. So everyone from Canonical to Red Hat to Suzy to SOC vendors like Qualcomm, Cavium, HiSilicon and then some of the end users like Alibaba or Facebook. They're all developing software together. They all develop software because they have a product or something that they rely on. So why even have Lenaro? This is what Lenaro does. We only work on the upstream. So there's a term that we'd like to use which is root upstream project. So a good example of a good root upstream project is kernel.org. It's the top level. And then underneath it, you might have, you know, Fedora that has one or Canonical or Debian. But we always work in the root upstream project itself. And so these are some of the ones that we're involved in today. And you can see you've got OpenStack, you've got Kubernetes, DPDK, and other ones. We let... Essentially, we're vendor neutral, right? So we just work on those areas. This side didn't come out, right? All right, so this conversation is supposed to be around containers. So I'm going to get into that now. So building the stack for ARM64, if you think about doing container work on Intel, you really, where you start is, you start with Docker, as an example, Docker engine, and then you start building on top of that, right? Everything's already there. You've got Golang is there. The Docker engine works. The Docker registry works. All of the different components. What we've had to do inside of Lenaro is make sure that all those core components are also working. So you can't start with a Docker container. First, you have to make sure that Golang is working. You know, you have to make sure that the kernel is enabled, that libvert's enabled. So all of these different areas that you have to sort of build up, even just to get started to actually work on either containers or open stack or some other workload. So what have we done so far? We've started this project really in earnest about two years ago. HEMA over here is the tech lead for software-defined infrastructure, right there in the front. And so we've done, what we've been able to do is actually provide a reference build. So you can go to the Lenaro website and you can download a reference build that runs on ARM. And we use it in production. Why did we go off and do this where we sort of went off of branch into what OpenStack was doing? Truth of the matter is, we couldn't get patches landed fast enough in the upstream. And in this case, it was OpenStack.org or in some other cases, it was KVM or QEMU or some of those. So we had to go ahead and build something. We still waited for the patches to land upstream, but we released it early. It is Ansible-based, so it's simple. We didn't want it to take a long time to actually go ahead and deploy. And it does relatively quickly now. It's packaged in both Debian and RPM. We had to be neutral, so both Debian and CentOS are supported. It's based on Newton. And we actually have three clouds deployed in this today, one in Austin, one in Cambridge, and one in Shanghai. And that's the developer cloud that Jeff talked about. So you could actually go and get space on there if you're developing against ARM. We'll give you an area and you can do your porting, you're testing your QA, everything that you need. So that's available. But there's a big problem with this. A couple of things. So then we've changed tactics. We ended up becoming package-packagers. So instead of actually focusing and working on innovating, working on ARM for OpenStack or some of the other different projects, we ended up doing packaging backports, backport packaging, forward porting. We had to QA both Debian and RPM. It didn't make a lot of sense to do this. We also ended up becoming maintainers as opposed to packages because we had our own packages that we had to carry with us. And so we'd have to watch as different packages changed, then we would have to go and change those. For example, in Debian, push them into backports. So again, we're not developing, we're not innovating, we're not working in the upstream. And it wasn't vendor agnostic. So we tested it on Debian, we tested it on CentOS, but that's not really what we're supposed to do. We're supposed to work on the open source effort. So what have we done? We've moved away, and I think everyone here is doing the same thing. We've moved away from packaging and we're going towards containers. So that's what we're doing. And so for part of that, what we've done is we're now part of the COLA project inside of OpenStack. And the work that we're doing is to try and make sure that everything that we do can then be reused by the rest of the community. So the biggest blueprint we have right now is to bring multi-art and start building ARM64 containers. That's the link down there. And there's a patch pending. So if you can actually influence that patch to get accepted, that would be great. We're doing this within Lenaro, but we have a lot of support as well from IBM, I believe, who also need this. But if you think about what you need for a container, if you're going to have something, you can't just make it just for x86, right? When you also have ARM and you also have power. So that's what we're working towards. So the target release for this is actually Pike. So this will come out pretty soon. It all depends on how long we have to wait for someone to approve that patch. We'll remind again. The COLA Ansible works well within the ARM64 containers. That's going well. The COLA Kubernetes, it's under heavy development right now. So we do think it will, do you think it's going to make it for Pike? No clue. All right. We hope so. That's the intent. But after this, so this is really where we're working with containers in the upstream project, being vendor-neutral, working for the ARM ecosystem to make sure that you have the availability of using ARM servers as opposed to just x86. All right. So where does this still hurt? There's a lack of... So in Docker, it's not aware of the different... In Docker, a Docker container, as far as Docker is concerned, Docker engine, the Docker registry, it thinks everything is x86. So it can't tell if it's an ARM image, or if it's a Power image, or if it's an x86 image. And that's a real challenge of trying to pull the containers in to actually do something. Especially if one container goes and pulls another container from a registry somewhere, then it can just go and pull an Intel one. So that's painful for us. So what does it mean? We have to manage our own registry and our own images, because that multi-architecture isn't there today. And we're trying to work with Docker to actually get that fixed, and that's one thing that's blocking both ARM and Power as well. COLA is a very inclusive project in OpenStack, and I think it's 198 packages of containers somewhere around there. It's a lot. We only QA and test a much smaller subset, because we're really sort of looking at the core reference of OpenStack and not all the other projects that are out there. For us, while we produce a lot of these containers, there's no QA or testing that's being done by them. We do QA and test what we use to be able to run a functional OpenStack cloud. And then the other thing is we keep on having developers ask us if they can now install OpenStack on their phone. And that's a true comment, right? It's not. It's for the... It's for cave-ins or Qualcomm servers, right? One or two or three, four-year systems. That is it. And I've gone through pretty quickly. We have three minutes. Any questions? Or do you want me to talk about other areas or challenges? The focus right now is around ARM64. That's really where we're doing all of that work. If you're talking about the guest inside of... So you have OpenStack and we're talking about all the hosts. If you're talking about a guest instance, depending on which chip you're using, which I think is almost all of them, that 32-bit guest will just work. There are some architecture designs where they don't have the 32-bitness in it. So then there's something called ILP32 that's able to manage that. Okay, yep, no problem. Any other questions? Okay, if there's no other questions, I just want to point out. So we did have a shiny object at the back there, the server board. Jeff is just wondering off with it, but the ARM booth is right there just behind you guys. So if you want to go and see, we've got a bunch of little boards all the way up to full-on server board. I'm happy to chat with you and get into more details or enter any questions you may have with that. Thank you all, I appreciate your time.