 Live from Manhattan, it's theCUBE, covering AWS Summit New York City 2017, brought to you by Amazon Web Services. Welcome back to the Big Apple. We continue our coverage here on theCUBE of AWS Summit 2017. We're at the Javits Center, we're in Midtown. A lot of hustle and bustle outside and inside here. Good buzz on the show floor with about 5,000 strong attending and some 20,000 registrants also for today's show. Along with Stu Miniman, I'm John Walls, so glad to have you here on theCUBE. And Dustin Kirkman now joins us, he's at Ubuntu, the product and strategy side of things, economical and Dustin, good to see you back on theCUBE. Thank you very much. All right, you just threw a big number out at us when we were talking off camera. I'll let you take it from there, but it shows you about the, I guess the presence, you might say, of Ubuntu and AWS, but that nexus is right now. That number. Yeah, Ubuntu easily leads as the operating system in Amazon. About 70%, 70% of all instances running in Amazon right now are running Ubuntu. And that's actually, despite the fact that Amazon have their own Amazon Linux and there are other Windows, REL, SUSE, Debian, Fedora, other alternatives, Ubuntu still represents seven out of 10 workloads in Amazon running right now. Yeah, huge number. So Dustin, maybe give us a little insight as to what kind of workloads you're seeing, how much of this was people that, Ubuntu has a great footprint, kind of everywhere and therefore it kind of moved there, and how much of it is new and interesting things, IoT and machine learning and everything like that, where you also have support. When you're talking about that many instances, there's quite a bit of both, right? So if you look at just EC2 and the two types of workloads, there are the long running workloads, the workloads that are up for many months, years in some cases. I've met a number of customers here this week that are running older versions of Ubuntu, like 1204, which are actually end of life. But as a customer of Canonical, we continue providing security updates. So we have a product called Extended Security Maintenance. There's over a million instances of Ubuntu 1204, which are already end of life, but Canonical can continue providing security updates, critical security updates. That's great for the long running workloads. The other thing that we do for long running workloads are kernel live patches. So we're able to actually fix vulnerabilities in the Linux kernel without rebooting, using entirely upstream and open source technology to do that. So for those workloads that stay up for months or years, the combination of Extended Security Maintenance, covering it for a very long time, and the kernel live patch, ensuring that you're able to patch those vulnerabilities without rebooting those systems, it's great for hosting providers and some enterprise workloads. Now on the flip side, you also see a lot of workloads that are spiky, right? Workloads that come and go in bursts. They maybe they run at night or in the morning or just whenever an event happens. We see a lot of Ubuntu running there. It's really a lot of that's focused on data and machine learning, artificial intelligence workloads that run in that sort of bursty manner. Okay, so it was interesting that when I hear you talk about some things that have been running for a bunch of years. On the other side of the spectrum, it's serverless and the new machine learning stuff where it tends to be there. What's canonical doing there? What's kind of exciting you? Any of the news, Macy, Glue, some of these other ones that came out, how much of those fit into conversations you're having and what's going on? Sure, they all really fit, you know? When we talk about what we're doing to tune Ubuntu for those machine learning workloads, it really starts with the kernel. So we actually have an AWS optimized Linux kernel. So we've taken the Ubuntu Linux kernel and we've tuned it working with the Amazon kernel engineers to ensure that we've carved out everything in that kernel that's not relevant inside of an Amazon data center and taken it out. And in doing so, we've actually made the kernel 15% smaller, which actually reduces the security footprint and the storage footprint of that kernel. That means smaller downloads, smaller updates. And we've made it boot 30% faster. We've done that by adding support, turning on, configuring on some parameters that enable virtualization or the VerdiO drivers, specifically the Amazon drivers to work really well. We've also removed things like floppy disk drives and Bluetooth drivers, which you'll never find in a virtual machine in Amazon. And when you take all of those things in aggregate and you remove that from the kernel, you end up with a much smaller, better, more efficient package. Okay, so that's a great starting point. The other piece is we've ensured that the latest and greatest graphics adapters, the GPUs, the GP GPUs from NVIDIA, that the experience on Ubuntu out of the box just works and works really well and well at scale. You'll find almost all machine learning workloads are drastically improved inside of GP GPU instances. And for the dollar, you're able to compute sometimes hundreds or thousands of times more efficiently than a pure CPU type workload. You're talking about machine learning, but on the artificial intelligence side of life, a lot of conversation about that in keynotes this morning, right, a lot of new services, whatever. Again, your activity in that and where that's going, you think, over the next 12, 16 months. Yeah, so artificial intelligence is a really nice place where we see a lot of Ubuntu, mainly because the nature of how AI is infiltrating our lives, it has these two sides. One side is at the edge and those are really fundamentally connected devices. And for every one of those billions of devices out there, there are necessarily connections to an instance in the cloud somewhere. So if we take just one example, right, an autonomous vehicle, that vehicle is connected to the internet. Sometimes, well, when you're at home, parked in the garage or parked at Whole Foods, right? But sometimes it's not. You're in the middle of the desert out in West Texas, right? That autonomous vehicle needs to have a lot of intelligence local to that vehicle. It gets downloaded opportunistically. And what gets downloaded are the results of that machine learning, the results of that artificial intelligence processing. So we heard in the key notes quite a bit about data modeling, right? Data modeling means putting a whole bunch of data into Amazon, which Amazon has made it really easy to do, you know, with things like snowball and so forth. Once that data is there, then the big GPGPU instances crunch that data and the result is actually a very tight, tightly compressed bit of insight that then gets fed to devices. So an autonomous vehicle that every single night gets a little bit better by tweaking its algorithms, when to break, when to change lanes, when to make a left turn safely or a right turn safely. Those are constantly being updated by all the data that we're feeding that. Now why I said that's important from an Ubuntu perspective is that we find Ubuntu on both, in both of those locations, right? So we open this by saying Ubuntu is the leading operating system inside of Amazon, representing 70% of those instances. Ubuntu is across the board right now in 100% of the autonomous vehicles that are running today. So Uber's autonomous vehicle, the Tesla vehicles, the Google vehicles, a number of others from other manufacturers are all running Ubuntu on the CPU. There's usually three CPUs in a smart car. The CPU that's running the autonomous driving engine is across the board running Ubuntu today. The fact that it's the same OS makes life quite nice for the developers. The developers who are writing that software that's crunching the numbers in the cloud and making the critical real-time decisions in the vehicle. And you talked about autonomous vehicles. I mean, you think about a car in general. I mean, thousands of data points, right? Coming in in continual real-time. Right. I mean, so it's just not autonomous. Right. You know, operations, right? I mean, are you working in that way, diagnostics, navigation, all those areas? Yeah, so we, what catches headlines are a lot of the hobbyist projects. The fun stuff coming out of universities or startup space. You know, drones and robots and vacuum cleaners, right? And there's a lot of Ubuntu running there. Anything from raspberry pies to smart appliances at home. But it's actually, I think really where those artificially intelligent systems are going to change our lives is in the industrial space. It's not the drone that some kids are flying around in the park. It's the drone that's surveying crops. That's coming to understand what areas of a field need more fertilizer or less water, right? And that's happening in an artificially intelligent way as smarter and smarter algorithms make its way onto those drones. It's less about the running Pandora and Spotify having to choose the right music for you when you're sitting in your car. Right. And a lot more about every taxi cab in the city taking data and analytics and understanding what's going on around them. It's a great way to detect traffic patterns, potentially threats of danger or something like that. That's far more industrial and less interesting than the fun stuff, you know, the fireworks. The shut-off fire drone. Not nearly as sexy, right? Yeah, it's just not as much fun. But that's where the business is, you know? That's right. Dustin, one of the things people have been looking at is how Amazon's really maturing their discussion of hybrid cloud. Now, you sit in data centers, public cloud, edge devices, lots of mobile. We talked about IoT and everything. What are you seeing from customers? What do you think we're going to see from Amazon going forward to build these hybrid architectures? How does that fit into autonomous vehicles and the like? So in the keynote, we saw a couple of organizations who were spotlighted as all in on Amazon. And that's great. And actually, almost all of those logos that are all in on Amazon are all in on Amazon on Ubuntu. And that's great. That's a very small number of logos compared to the number of organizations out there that are actually hybrid. Hybrid is certainly a ramp to being all in, but for quite a bit of the industry, that's the journey and the destination to, in fact. That there's always going to be some amount of compute that happens local and some amount of compute that happens in the cloud. Ubuntu helps provide an important portability layer. Knowing that something runs well on Ubuntu locally, it's going to run well on Ubuntu in Amazon. Or vice versa. The fact that it runs well in Amazon, it will also run well on Ubuntu locally. Now we have a support. Yeah, I was just curious. You talked about some of the optimizations you made for AWS. Right. Is that now finding its way into other environments or do we have a little bit of a fork? We do. It does find its way back into other environments. So the Amazon hypervisors are usually Zen based, although there are some interesting other things coming from Amazon there. Typically what we find on-prem is usually more KVM or VMWare based. Now most of what goes into that virtual kernel that we build for Amazon actually applies to the virtual kernel that we built for Ubuntu that runs in Zen and VMware and KVM. There's some subtle differences. Some, a few things that we've done very specifically for Amazon, but for the most part it's perfectly compatible. All the way back to the virtual machines that you would run on-prem. Well Dustin, always a pleasure. Yeah. To have you here on theCUBE. Thank you John. You're welcome back any time. All right. We appreciate the time and wish you the best of luck here the rest of the day too. Great. Good deal. Thank you. Glad to be with us. Dustin Kirkland from Canonical, joining us here on theCUBE. Back with more from AWS Summit 2017 here in New York City, right after this.