 So we had this super computing event here in Denver and you're wearing a red hat right now this is a red hat day what's going on? Yeah so I'm John Masters, Chief Arm Architect at Red Hat and today's kind of an important day for us here at Red Hat because we just shipped Red Hat Enterprise Linux for ARM after seven years of research and development and activity that's led up to this day so it's a huge milestone for us and it's not the end it's the beginning of a multi-year journey that we call our multi-architecture future so we work across many different architectures at this point and ARM is one of them. So Red Hat Enterprise Linux launched, what does it mean to have all this launched? Well what does it bring? So Red Hat Enterprise Linux is our flagship Linux distribution that targets the enterprise and what it does is it brings a common experience across many different architectures that customers choose to run so if you install Red Hat Enterprise Linux for ARM you're getting the same experience that you would get say on you know one of the incumbent architectures like x86 whether that's Intel or AMD or power or or system Z for mainframe so you get the same you get the same experience yeah that's the that's the key so if you look over here there was maybe you can go back one slide so you are standing right there with the Qualcomm Centric 2400 what are you running on? Yeah so there's a couple of things we're showcasing in the booth here we are showcasing a system we launched today which is actually an HPE system that we launched in collaboration with Cavium and it's called the Apollo 70 and what it is we'll come around here and I'll show you what this is is a two socket yeah I will yeah this is a this is a two socket 64-bit ARM server so when I say two sockets what I mean is these these are this is one machine so it's not like an embedded device where you know these might be two different computers or something like that these are two sockets of one machine so each socket has up to 32 cores and each processor each socket has up to four threads per core so in in one configuration that could be up to 128 threads per socket times two so 256 threads if you want to configure it that way now in HPC a lot of people don't don't configure their machines that way they they really want to get the maximum single threaded performance so they'll probably turn off the multi-threading feature and they'll instead go with the 28 or 32 core configuration but then you double that so you're looking at 56 or 64 cores that they can use in one machine and these are very very high-end cores these are comparable to the other architectures out there whether that's 686 or other architectures so in fact this particular platform the selling point is not only is it very close to the performance you see on other architectures in terms of memory throughput it's actually 33% better than the the incumbent solutions so there's a compelling reason for customers to be interested and that's what drove our launch today as HPE announced the Apollo 70 we also co-launched the Enterprise Linux product and now as I mentioned before the Enterprise Linux product targets 10 different I think I hinted this to you separately but this this product targets 10 different micro architectures so in fact more over time but right now it runs on 10 different implementations of the ARM architecture and that's for many many different vendors so companies like Cavium here and then also other companies like Qualcomm and the video I was playing in the background here is actually a demonstration well I guess a little repeat in a second the video you're going to see behind me is actually taken from the Qualcomm Centric product launch last week and Centric is targeting a similar set of market customers and what we were showcasing with with with Qualcomm was that our ARM story is really a multi-vendor story right whether it's Cavium whether it's Qualcomm whether it's the company that was formerly may come or applied micro and now is called Denver project holdings it doesn't really matter which company you're working with because you're going to get the same experience and each company will have an opportunity to differentiate and add value of their own and so Cavium with this two sockets system here is is really bringing a certain set of characteristics and then Qualcomm with their single socket solution is targeting a similar set of customers maybe with different workloads and what we did at the launch last week there is we showcase this technology we have called RDO which is used to stand for Red Hat distribution of OpenStack and now it's a it really just goes by RDO and it's a community project to enable OpenStack across different architectures and in a Red Hat distribution friendly way so it runs on Fedora it runs on rel it runs on other compatible operating systems and what I was doing here is I was showcasing that when you're on an ARM machine this happened to be a century it could have been a Thunder X2 the Cavium platform but when you're on one of these ARM server platforms not only can you run rel but then you can start to enable the other technologies so here we're running a an OpenStack cluster in a box and I was showcasing that you know here you can see this video I'm going into the the management console here and I'm just showcasing you see by click instances here and then I say rel7 demo and I say you know I want to create a rel7 instance and I click on rel7 for okay and I just I just launch it and then in a few moments the video will show you show me you know logging into that machine and configuring it and here we go I'm going to ping it this VM I'm going to configure it and play with this instance and then you'll see behind me you'll see some more of that in a moment but the the fundamental point is now that we have rel it's the enabler to all of the layered technologies that we have as a Red Hat portfolio and so you're going to see Red Hat engaged in all these communities enabling technologies like RDO and then as these technologies get enabled when there's the customer demand and the right market conditions then they can become products and that's always based on what our customers are asking for and the timing the customers are asking for because you can't ship everything on one day but as I said we do have builds of all of these layered technologies that run on top of a rel foundation across every architecture so this is a work you've been working on this stuff we've been doing videos for a bunch of years yeah I've wanted to make this video for a long time and for example HP has done the moon shot and then they were doing something before Apollo also yes yes so everything is leading up to something yes something is happening something special right now I think so I think the the the problem we have in the industry is that well the challenge we have is that everybody's optimized around a common solution they already have the timeline yeah yeah you can jump into the timeline so yeah you can see here the you know 2011 we start the team in March and then we have a series of tech technology milestones here as we work on on our enterprise group our early access program dev previews leading up to our product launch and so you asked me about you know what sort of why have we had these different technologies up to this day and and how is it that we we we've taken so long perhaps to to to see arm server adoption at at scale and I think the reason is that the industry is optimized around a certain way of operating that it has for years so you know a two socket system with a certain level of performance and so on and I believe for a long time that we have to give customers that same experience with arm really as a as a as a as you know planting a planting a flag and saying you know this technology is ready and you can run your workloads on it and that's something I've pushed the arm vendors to do because it's not my place to go and create the market but I give them feedback on on how I see technology being used and I think that requires a certain level of performance and it requires a certain level of capability that we didn't see there in some of the early platforms now perversely I think there's an opportunity in the future for some of those other concepts to really take hold now that we have a two socket server system that's at a high level of performance that's got more than a terabyte of memory and all of these features we're able to go back to some of these folks and say well sure you can run that if you want but maybe you don't actually need to have that high level of performance because maybe all you're trying to do is attach a workload accelerator to your machine a GPU or something else and really old maybe all it's doing is spending its time marshalling data to and from the CPU or rather to and from memory CPU may not be doing much work and so these platforms that we've announced over the past week or the market has announced and then that run rel what they do is they show the industry that arm is not a toy arm is a mainstream server it's able to run real workloads and then what you're going to see is you're going to see a level of adoption I think as people realize okay I could run a workload on this and then they're going to look at it and they're going to say well maybe I don't need to run every workload on a high end core maybe I can run it on a smaller core so maybe the technologies that you mentioned in the past like moonshot I think they were just ahead of their time they're not a bad idea in fact they're a good idea it varies really depending on the the the environment and I'll give you one more example if I'm building at scale I may actually look at a technology that's kind of win-peer core like you mentioned the moonshot platform which I think is a good platform and you might look at it and say well I have I have all these different cartridges and each one has a lot of cost associated with having memory and other features and I may get a cost saving by simply going for that two socket model because that's how I've built all of my existing infrastructure so I think that's what we're going to see first but then over time as some of these new technologies take hold you're going to see ways that we can in some sense return to some of these other ideas all right so so you're giving people a lot and letting them scale down exactly I think that's how the industry works because as I said the industry is is used to working one way and so we have to give it what it what it expects and what it thinks it wants and then what I think is going to happen over time you know the the joke with Apple was always that you know Apple would give you exactly what you know they thought you wanted right well they're in a position in the market where they're able to do this with technologies I think the arm vendors are not yet there so what they have to do is establish a beach head and establish themselves in the market as you know not wimpy cell phone processors but really high-end capable machines and they're going to do that with some of these designs they're all able to run the same OS so from a red hat point of view our customers were able to have the same experience whether they choose to adopt on arm or x86 or power or you know system Z the main frame and in fact what you're going to see over time is you're going to see new market potential because the purpose of this is not to be subtractive or simply to replace you know a power or an x86 or a mainframe machine there wouldn't really be much of a point to that the reason we're involved here is we see opportunity or our customers rather see opportunity in new emerging markets so HBC is always an interesting space because everyone is looking for innovative designs you know Moore's law is basically dying and as a result you're not getting that performance increase year on year so the the national labs and others in the the HBC ecosystem are saying you know what we're going to do is we're going to try arm because we think that gives us a path in addition to the other technologies out there to try new and novel approaches but then what you're going to see as a result of the traction the arm is getting is I think greater adoption in other areas that are new business growth for red hat that includes edge computing that includes IOT and that includes telco and we see a lot of interest from the telco community right now they'll be to take platforms like this one maybe not optimized for a telco environment but they'll be able to take these mainstream platforms that can run well and they'll be able to build some very interesting solutions with that and then they'll be able to go back to these OEMs and these vendors and maybe drive some solutions that are more tailored to the the problem space that they're operating in and what's the business model at the Red Hat how does it work so Red Hat's business model is very simple it's a subscription based model so customers do not buy rel customers buy a subscription and then they're able to choose which version of the operating system they want to install now on arm it's fairly straightforward because there is only the version that we launched today but over time you could see a model in which it's essentially similar to the other architectures where a customer has a subscription that means they have access to the Red Hat technologies and then they choose the release amongst an available set of releases that they want to install based on their environment so for example on x86 we have customers that are not even yet running rel 7 because they've got established setups that are running and they don't want to disturb them and so the part of the value of the subscription is that they're able to get support on older versions of the operating system and the same thing would be true here over time but it's key to point out that customers don't pay for rel as buying a boxed product customers buy into a collaborative model in which they pay a subscription and what they get is this kind of investment that drives a lot of the technologies that we're building here right support it's exactly you're you're you could you could look at it that you're taking out the insurance policy right we build some wonderful technologies but you know you've seen over the past few years you've seen you know some some security exploits have happened in the industry and and other kinds of issues that come up and what red hat is is the part of the value of red hat subscriptions is that we get ahead of all of these things and we address them ahead of time and we roll them out to customers in a way that's consumable when they need it and the other thing that we do actually is we work very heavily in upstreams so I mentioned technologies like open stack and staff and and even virtualization things that we don't yet have a comprehensive portfolio for on the arm architecture but what's happening is we're engineering these and we're taking those that the investment that's coming in through subscriptions and we're driving the upstream development of open source and then as a result of that we're able to offer that to customers in a way that's consumable consumable by them because a customer really doesn't want to sit there and figure out every single way to build open stack and get it working what a customer wants is to come to us and say you guys are the experts you make this work for me because I've got a workload I actually want to run I don't really care about the rest of it I want you to solve that problem for me and that's the value that red hat offers by getting upstream by developing these technologies as we have done for the last seven years on arm that's led to rel today and then over time we will enable these layered offerings that bring even more value to customers and all these guys that red hat the sum of them here right yeah top-end guys that you work working with some of them yeah that's right I think I can I trying to point out some of my colleagues here but there's a gentleman here in a suit with glasses Jan Fisher he's in our product marketing organization and he owns the multi-architecture product marketing so he's been involved with the arm offering and also with the power nine release we had last week because we do have a multi-architecture strategy now so it's not just about you know arm it's not just about x86 it's really about having a multi-architecture future so Jan's been very heavily involved and very very grateful for the work he's put in and he certainly has helped to organize this year supercomputing presence which I think is going very well we have a great virtual reality demo here as well and more beyond beyond those of us who are here we've got some folks here for you know public sector government customers and all the kinds of folks you would expect to be at supercomputing we also have a very large team of heroes inside red hat who have made this happen and one of them for example Brendan Conaboy he and I were the original creators of the arm team so I was the one pushing to make the team and then what what happened was my VP Tim who has been amazingly supportive the whole time and really it does take that kind of support to to let you do this for seven years you know Tim I Tim said to me in his office I showed him a beagle board actually one the small arm can embedded boards and I told him it was a small 32-bit board I said this will be a server one day we went through all the reasons why we'll get there and we need to be a 64 bit architecture because we don't care about 32 bit we need standardized platforms all the things that we eventually built together with Arm and the others but I'm walking Tim through this and he said well this is great technology sounds good you've got the the right ideas here this this could go somewhere let's let's do it but he said John you know your time optimistic you know I'm a I'm a good technologist I think but sometimes I I imagine that amazing amounts of stuff can be done in in very little time and so my colleague Brendan was brought in to help me with that and Brendan is a fabulous technical guy he really understands what's going on but even more than that he really understands from having run many projects in our embedded engineering organization he really understood how to drive things to completion over time so I would come up with these sort of technical ideas and concepts and he would then help us to execute those with the team that we had we had a small team to start with so it was very crucial to have someone to kind of be the task master and keep things going and then over time it's expanded so now there are not not just ten people working on arm as it was in early 2011 but now there are I like to say 10,000 people working on arm inside Red Hat because everyone's involved in arm on some level and in fact every time we build software inside Red Hat it is built for every architecture at the same time so there's there's no longer any notion of an architecture being special every architecture is the same internally and so so and you've always been pushing arm pushing the industry pushing the narrow everybody towards this goal this vision of the meeting you had seven years ago maybe kind of like it's important role yeah well I wrote a an email in 2011 and in that email to the team is a there's a loud speaker behind me so I'm gonna just repeat this in a second actually when I'm okay all right well I'll say it again so so we start the team in March March first 2011 and so when we start the team I sent a welcome email to everybody and I realized what I thought would be the things we'd have to solve so standardized platform standardized boot architecture standardized SOC's all the things that Red Hat needs because we can't play with arm as an embedded business model that's not what we do we have to have one common operating system that runs across every system so I itemized all the sorts of things that we would need to solve and then we worked out how are we going to do it and so this is what led to the creation of groups like the enterprise group inside lanaro and the enterprise group inside lanaro we started by four of us myself and three friends from other companies HP arm and and one other and we got together and we basically drove the creation of these different industry efforts and then for example the some of the server standards you know I wrote the first version of the SBBR which is the boot requirements from servers myself and now arm has a whole bunch of people that work on this and some great friends at army there's some amazing friends of mine at arm there they work on evolving the SPSA and they've actually recently announced a program called server ready which is a validation program that really allows you to say something is an arm server and not an embedded device so we came up with a lot of concepts early on for how this would play out over time and what I do every year actually is I send this email that I sent to the team in 2011 I send it out as a kind of happy birthday email and we go through kind of where we are and actually it's amazing how you know I'm not a I don't think I have a magic perception of how the industry works but I think if you've been around long enough you can tell the sorts of things that are going to be needed and it's been quite interesting to see how it's played out I haven't been a hundred percent right but I've been very close on a lot of things because you know we know how enterprise works we know how we have to address it and we've been able to give some really strong guidance to the arm vendors it's up to them to go and build the technology and make the market but we've been very very engaged and you know I regard them all as family in that sense and I'll tell you one other thing that's been very very rewarding for me and I think for the company is co-engineering so on some other architectures I've seen in the industry you know kind of the engineering happens within one company and then they engage upstream and then they get engaged with the Linux vendors and so on and that works pretty well but in the arm case what we've been able to do is get involved during the silicon design and during the early days of the design of these platforms even during the bring up so for every single arm server chip you you've ever seen pretty much I was there in the lab when it first came back from the fab helping with the bring up helping to get operating system software running of course standard Linux kernels and so on but also making sure that rel was running very early on and really building up a relationship with all these companies and it culminated in this program we have with HPE called the Comanche program the Comanche collaborators that was our code name for the engineering platform that became the system that HPE announced today the Apollo 70 but the the collaboration that we've had with HPE has been a little bit unique compared to some other platforms where we've actually worked in partnership with HPE with Melanox with the national labs we've all had very early silicon we've all been working out the bugs together and what it means is we're building a much more capable platform as a result because by the time it ships we have done a lot of battle testing ahead of time and right here some beautiful hardware let's have let's have the whole arm service thing happen right now you know it's seven years of my life as I was saying a lot today it's been a similar amount of time for a lot of people actually in the industry and we always thought it would be a ten-year journey so what I'm looking forward to now now that rel's there is building the next wave and the next wave is when it gets really interesting customers building deployments customers having good experiences and then customers asking for higher layered products that go on top of the rel foundation