 Hi everyone, can anyone hear me? Okay, I'll shout at you a little bit So first I have to apologize so I'm I'm Revamping this talk. So you're we're doing an experiment you might guinea pigs For a new version of this talk where I'm shifting a little bit the focus I have material for about two hours. I'm showing every other slide and we'll we'll see how we get through that Quick introduction, Daniel Rieck is my name work at reted in the CTO office. I've been in Linux forever Did a startup in the late 90s and I was the PM for rel for you know a long time So a lot of experience with the operating system. I was involved in the Container strategy in reted now I focus more in machine learning, but I'll I'll say how that's relevant The title of the gray beards was nightmare originally was inspired by by the the Debian fork over system D and The what happened is that system D changed how? Linux works right. It's a pretty big move away from Unix tradition to more actively managed operating system and That was for for some people in the Debian community that was too much change And there was a fork of Debian when Debian accepted system I think that that is gone away in the meantime and everyone has accepted to live with system D Some of the things that we are doing to the operating system right now are bigger changes and you know So Linux is changing more than system D did and so you know the the idea was I was this is This is even worse. So let's see what happens if we introduce Kubernetes into the mix So what's a traditional role of the operating system right? They are there two views traditionally that you can find in in in reted definitely or in general right? There is the infrastructure up view. We say oh, it's a hardware centric view right? we have an operating system in order to bring up hardware and It ties you know the operating system. It's very kernel centric, right? That's what talks to the hardware and the purpose of the operating system is to make my hardware work There's another view which is the application centric view which is a top-down view in the user space century It's the oh the operating system is what gives me a runtime environment for my application and Well, of course the truth is it's both right the whole point of an operating system is to allow you to run applications on hardware or something like it, right? It's not hardware any more necessarily It could be a virtual machine But it is there's always an intrinsic conflict between the two views because you know One reason like do I tie my operating system lifecycle to the hardware? Do I tie it to the opera to the application lifecycle and what's the ultimate purpose and You know traditionally You know back in the day This wasn't a big question because you had kind of a vertically integrated Combination of hardware operating system application right if you go back to the main frame. It was basically Here in the IBM model at least to some degree black box application running on leased hardware Right, so you you you had access to an application was very well integrated. It did what it needed to do But you could never take the application to a different piece of hardware Right, or you couldn't change the use of the hardware. You couldn't control how you use the hardware. It was vertically integrated You put data in and you never got your data out, right? It was like it was a one-way Roll and even in Unix so-called open systems. You still had the integration of hardware and operating system Right, so you know if you had a Sun server you had Sun hardware and a Sun operating system And that was true for all the major Unix what linux did the role of linux was to break up this vertical integration Right with linux. We got the ability to choose Hardware that we like from multiple vendors One little sort of and then choose software from an open from an open ecosystem, right and That linux wasn't the only operating system that had some choice windows had some of that Right, but even in windows you had The the you had the integration of one vendor with the operating system Microsoft itself Right, and they controlled to some degree still what you could do, right? And they controlled the tool chains. They controlled what ISVs could do to certain degree linux is the only thing that gave you full at Large you know in the mass market is a full free choice both on the hardware and on the application and provided a binary portability To the degree that's possible Based on the hardware limitation, right? So it you know you can take one One application on the source code level across different hardware architectures on the binary level within one hardware architecture You can take the application run it everywhere And that is still the role of linux today, right? It's a common binary runtime across Across hardware from different vendors across virtualization and across the public cloud That's why by everything the public cloud runs linux basically nowadays, you know because it gives you this abstraction and so You know so that you know that's always been what linux did and you know There's the the traditional of I'm running I'm running one machine, but we're seeing some shifting Paradigm some some big trends, right? And so one one thing we're seeing is that everything is software today, right? There is no business at large that doesn't depend on software one way or the other right even you know any kind of Any kind of what product like you know a car today is a data center on wheels Right, and you don't even have to go to like self-driving cars for that like any car has a whole bunch of computers in there And the functionality of the car is defined by software There is you know in the in the whole Electric car move with Tesla, right? What what? Tesla stated a point is that there Their challenge or their competitive advantage wasn't the car itself Right, it was what's the electrical piece, but that is a software to run the car, right? That's where the differentiation between cars happens nowadays So that shows you like the software is eating the world, right? Everything is software every business is driven Business value is driven by software And we have a fundamental change in how we how how we interact with With things right so It's like when I grew up where we lived in the broadcast model, right? So they were like schedules when things happened, right and you watch TV You know when a show was on right and you listen to radio where you had no control over the schedule, right? so it was like it was a life where You know what you could consume and how you could use things was Relatively like controlled by others outside of your control that has completely changed right nowadays our expectation is that is Things are always available when we want them, right the Epification of the world and that has not stopped it says that that has happened in consumer Things, but it also has happened in the software industry, right and modern Because of everything depending on software Model where like software is a one-way route and doesn't work anymore, right? And most lines of business has their own software developers, right? So they're creating their own software and their expectation is that the value they want To put in their business, you know the business the the requirements to create the business values I want to create in software are always available, right? So it's it's basically an application. It's a consumer centric idea so software Software developer if you go to modern software developer in a line of business who is pressure to create business value for their company Which is not a software company in the traditional sense, right? It's some company Building some consumer device They will not when you go to them and say oh you have to standardize on this Version of a library that has this limited functionality while there is some newer version somewhere available They will not understand what you're trying to do. They will not follow your standardization They're going to try to consume the newest version of something that has the features that enable their business value, right? so that's an important from for the operating system and the software stack right because You know traditionally we tried to standardize things and that has changed fundamentally You know, there are other things You know the complexity of software stacks has grown so we see an aggregation of services rather than monolithic systems And and you know and well what's nice for us is open source has become the defaults for software, right? the The differentiation of proprietors, so they're still proprietors often is of course always going back and forth But at the core of most software systems today is open source. So that's the foundation for for software that's That's become and customers consume it that way and then the stand open source. So we have a lot of Interactive is whether any of some of you probably, you know I'm talking a little bit from a retic point of view because that's my world but you know many of you probably are in in non-software companies and You understand open source you interact with open source even though you're not in a traditional software vendor Right, that's a fundamental change and that's very common in the cut with the customers that that I interact with that They are advanced users of open source that actively contribute to open source actively participate in open source which Which changes the dynamic, right? It's go it goes back to like it. It also makes open source From emulating a proprietary broadcast model of software, right? open source business now has to and and open source systems have to understand this value cycle of interaction with Everyone in the value chain also contributing back to open source There are other changes move to cloud native behavior And I'll talk a little bit about that is a big change deaf ops agility those kind of months are changing how we treat software A big piece You know in all of that is software stack complexity and you know, I put this yesterday from module counts It's from the internet, so it must be true But you know it like even if it's of like, you know, so this is just a number of of projects in common Repository and these could all be for like forks of the same project But that's that's basically what the line of business developer pulls from right and You know if you see like npm is this like explosion, right? and That's that's a really interesting problem, right? So so I think in Fedora we have a Reader of about 30,000 packages and Debian has slightly more So that would be like somewhere somewhere down here, right that what we can package as binary packages So there's no chance that that for example linux distribution is traditional sense can package all of this and Make it available in the way we have we make available the core system, so and so that that's an interesting Change in how software life cycles work And like and so this npm stuff is really interesting. Well, this is not just no jazz that might actually that's a good question So there's a lot of code in there that actually like isn't executed on the server Right these are frameworks that put things that are executed in your web browser. I hope my presentation is basically it's right. It's Google Slides I'm using right and that's basically a JavaScript app that's running in my browser That's pulling code. Well with Google. They have pretty good controls over but You know in the in the traditional linux model what we usually see is that? You you package software and it's on a server and the security problems are confined to the server And which can have significant impact if that's insecure, right? But what we're seeing now is like the way that modern software works is that the browser itself downloads additional software packages when you execute an application and so like so so the whole like concept of Still of like software main stability and and security like has a much bigger implication now it's much more Complex than what we traditionally had to deal with In in traditional software that which is like a server with a terminal or a local machine a workstation And so and you know the I call that app Send up centricity, right? So the one of the problems we're dealing with in linux I know I said the there's like a diminishing return on the you know against the complexity of software stacks I mean because packaging everything in the linux distribution works well for the core system Right, basically the kernel g lip see and the pieces you need virtualization the pieces you need around But it doesn't it can't keep up with the complexity and even the Use and deployment model of these complex application stacks today. It just you can't can't keep up and You you know everyone who develops software will go to the native package format for the software stacks They're using and which in npm with these npm packages then in the browser application It's downloading random stuff sometimes and like there's some it gets really scary what what if you go like to Agency created or line of business created software And you watch what what it's downloading. No one knows what's in there anymore, right? Like so no one can tell you where this stuff is coming from what it does in in most cases, right and You know in go it's all source code and then You know that gets compiled into static binary, so like there's no way to know what's in there like once you get it Right, so they're the interesting problems. They were where like what we did traditionally was packaging Code in an rpm for binary distribution, right? Which was a huge step forward against the fragility of compiling things on that each machine But it it's not sufficient anymore to keep up with this because you can't capture all of it another big change is That you know In the modern world like it the cluster is a computer I see the whole concept of deploying an individual machine is not a reality anymore. No one no one runs one machine, right if Sometimes you still look at an individual machine, but that's usually a virtual machine Right. I'm running on a cluster of virtual machines or in the cloud in most cases You basically look at services that that that You provide so it's it's basically One service that gets deployed somewhere that service might be implemented as a VM or as a container progressively it's in containers and with you know with in Linux, that's where where Kubernetes comes in right so the Problem we have You know traditionally everything we have done in the Linux distribution is confined to one machine, right? That's how we looked at it and so You you had the special case of clustering so so system. These are single machine Orchestrators that takes care of your machine Really well, but then you know when you cluster them that becomes a special case and you have some failover But there's nothing in there that takes care of like how how do services move? Across multiple machines. How do you get connectivity, right? And that's where Kubernetes comes in and Kubernetes now has become the default orchestrator for services on Linux That run run in in clustered services The It inherits the the features from Docker and or like Docker like OCI containers Which I'm sorry, I have to look at the time All right, so so it inherit its built on The container concept those introduced by Docker first standard as an OCI now implemented with potman and and and cryo in in red heads stack and What it does basically you can look at it when we sometimes call it a meta kernel, which of course is More meta meta system D right it it orchestrates services across Multiple machines and manages the cluster and then on top of that it services that that orchestrate how you talk to the service direction service mesh how calls get routed in that server and What's What's interesting is that is that it Does two things so the container concept encapsulates the dependency stack of an application Into a namespace Which removes one of the biggest problems that that you have in the traditional model of of an operating system, which you know You can call dependency hell, right? So if you have in the traditional single node operating system, you have one user space, right? And when you want to install things they have to They cannot conflict in that user space, right? That's how we our PM works, right? I PM solves that with its dependency management But when you go into this model of many different package sources that gets increasingly complicated, right most of them have some way of isolating like a virtual environments, but The dependencies on the underlying system then still need to be Consistent or interdependent between these systems and so traditionally that has been solved by running thing in virtual machine So you could do one virtual machine per service, but that's relatively expensive So containers of a very nice light-weight way to give you a user space isolation to manage one service Dependent binary dependency stack in a nice way and then Kubernetes gives you a way to orchestrate these together to serve applications out of a cluster, right so Bunch of things happening here right move to from one server to a cluster of machines and from kind of the binary like one application binary interface to the operating system to abstracting that and splitting the operating system into the hardware piece and An application piece that lives inside the container and and can move around so so the Equivalent to the earlier picture, right traditional in the traditional server We had red head enterprise Linux between the application and the infrastructure in the hardware or the virtual machine and now in the new app-centric platform we have Redhead OpenShift, which includes enterprise Linux, right? That's still that's still the foundation for that So it's enter president press Kubernetes plus a bunch of other servers that becomes a new abstraction layer for applications across the different footprints of infrastructure now This is happening in the context of cloud, right? And I talked a little bit earlier how cloud is changing how we do software, so to go a bit a bit deeper so the cloud It's not only that That clusters are the default the default for most large customers today are cloud deployments for new projects If you're not in an regulated industry Then you default for new projects to the cloud in many areas and There are a bunch of reasons for that right one reason is OPEX versus capex, which I think is The least important one. It's it's like it's from a budgeting point of view It's easier to go to the cloud because you don't have to buy hardware with a long commitment The key points are elasticity and self-serve you can just go there and just do it It's much easier to get the cloud instance than to go to internal process even to get an internal VM in many companies I just go there with your credit card and then the companies have to deal with that You can move up and down as you need right if you buy hard with the traditions and moving down isn't really an option You already bought the hardware cloud gives you the ability to reduce and be elastic the cloud has managed to Establish data aggregation all your data and and progressively today You know, it's a separate topic that that we have all thread on is you know the concept of intelligent applications applications that Will bend and define that for me applications that are learned from data and improve their behavior based on data Right those applications need to live with data or need data access Cloud allows you to aggregate data very easily and so intelligent application have a high affinity to cloud They are the service of cloud provides great integration of different services They often have better security right so you can trust that Amazon's infrastructure is very secure It's very expensive if you run your own infrastructure in a public-facing way to keep it as secure as they will keep it For a bunch of reasons It's their core business, it's probably not your core business And they have operational excellence right they they because it's a core business to operate that cloud They know how to operate that cloud no individual company can compete with that directly They run everything as a service right so they provide pre defined services They are now moving on prems. It was a hybrid model So even if you have things there you say, oh, I need to keep certain data or certain processing locally for latency for example, they have For industrial IOT for example, they have on-prem Satellites now that you put in your in your factory and they're basically part of the cloud But local data is processed locally so they move hybrid which you know removes a lot Is very attractive for large users that needs the data aggregation to central views integration and You know the key from my point of view the key piece what they're doing is the service abstraction and time to value in the public cloud right that you interact not you You can choose your entry level right traditionally like it was VMs, but now it's services All right, so you aggregate your data your application from a bunch of predefined services And then you add the piece that that's different for you So you can focus purely on your own differentiation And you don't have to build all the other pieces right and and that's You know you can focus on your core value and I think you know the concept that like shows that the most is function as a service so Amazon Lambda Where the platform like you basically only deploy code snippets in a predefined framework You know usually for things that act on triggers on data triggers and messaging triggers So you deploy only like you basically you don't have to you don't even have to deal with the operating system Was a software stack anymore as a runtime right? It's all predefined That way abstracted and then it talks to it through a messaging bus through some other predefined applications So you know that the the that gives you this level of service abstraction You don't need to be an expert in any of these other things right you don't need to storage admin You don't need an DBA you don't need To you don't need sys admins to install messaging you can just use that now in an environment where your data is already there and They give you this on-premise thing to aggregate your data and and push some of these decisions locally. That's a very compelling Value proposition, right and and that has changed how we look at software in general right like so You know traditionally it was what we call enterprise software So someone writes software that someone else then Deploys and the third person uses right now with with you know everything being software line of business having their own developers we are do people just writing applications on top of a of a stack that is completely controlled and provided by third party And and so so there's a big downside to the whole thing And that is the lock-in. It's black box services. You have life cycle dependency data gravity You can't get your data out once it's there or it's very expensive to get it out You have no reproducibility because you don't know what these black box services actually do right You can't look inside you if they change how it behaves You you have no control In you know, it's really bad for open source and you know probably a lot of you have have seen some of the Fights of our licensing where like open source project or the companies that built an open source project Are changing their licenses to make it harder for the cloud vendors to offer their software as a service Because there is no room for ISVs anymore in this model right Amazon It's Amazon basically competes with the whole rest of the software industry at the end and and so You know, I said it's bad for open source. So who cares right? Why why does it matter if it's good for customers? Well, the problem is that they're there good reasons why you want open source, right? It's a transparency It's long term strategic independence. It's innovation, right? If you go into monolithic system You basically they basically recreated the mainframe, right? It's black box services running on leased hardware That's that's what they did. It has all the advantages that the that in the early days of computing the mainframe had and You know, you can just use it. You don't have to deal with it. It's very secure very reliable But it has also all the downsides. You're totally locked in you can't innovate you can't differentiate you can't do things differently and You basically it becomes more expensive because you're in it in this dependency, right? So so And and you know a big conflict is like customers who want to have strategic independence We don't want to buy into this monolithic system and a choice of like four big vendors that are relevant, right right now For all kinds of reasons And the software vendors that are not one of the four Big ones have a shared interest to find an alternative to them And what you know, what are the challenges for that alternative? The big challenges is is the service abstraction time-to-value. That's the first one right if you like if if I go to a customer and say, okay You want you know, I want I want to enable you to run things offline on premise, right? Not in the cloud a service. Well, you need to install Linux and then you need to install Your database and then you need to install this and this and this and I end up with this huge software stack that they have to maintain They have to become experts in You know, they need to deal with storage. They need to deal with hardware. That's sometimes that's a very tough Value proposition, right? The sheer number of servers are easily available, right? I can just click a button and I can just deploy a service and I it covers my whole application architecture That's late. That's a lot of work if you want to do that in the traditional model of you know, how we bought software traditionally right Because you have to deal with five different vendors to get to the same in Amazon It's just one vendor you need to do it. So that you know, that's that's a problem Application portability right like if you go like can I can my application? Move between the the cloud right so so one of the big advantages of a third-party Model over a single cloud vendor is that if you know If you can take your application move from one cloud to the other cloud or on prem Then you immediately are in a better negotiation position against the one vendor, right? So we need to provide Application portability and finally we need to have the operational excellence We need to make the operational excellence available to customers that Amazon provides as a benchmark was the other club I don't want to single them out But but we need to make that available in a model where as the average IT company Can can or the average IT department in an average company can operate their IT in a Competitive manner right which is really a hard challenge because you know The expertise is never going to be there And in a way, you know, you can look at redhead's traditional business model as making Make using Linux to allow companies who couldn't Who couldn't afford a Sun server to run the IT like they could right that was in the early days was our business model So it was about enabling people at a lower cost on PC hardware to do the same things that that you know the pros did in in with Unix service You know at the end like now our challenge if we need to enable the average IT department to run their IT like Amazon does right So how do we do that? so we think that Kubernetes is The answer here right as I explained earlier Kubernetes is a service orchestrator in a cluster it provides you OpenShift is redhead's distribution of Kubernetes, so I use in In synonym here, although like technically Kubernetes is kind of the kernel of OpenShift and then the other services around so OpenShift gives you you know and extends Linux to give you kind of a service application-centric deployment model across clusters, right? abstracts from the underlying hardware it takes care of connectivity of Applications in that cluster right between the application how you get to the application it has application lifecycle management So it has it's a cloud platform in that sense that it gives you this Elasticity and service-centric or application-centric view Applications are portable because they're defined independently of the underlying infrastructure right and even So the portability can happen again like in traditional Linux on two levels Like if you have the same hardware architecture It will inherit the binary compatibility and you can take your your multi-service application definition and deploy it in parallel to multiple places I think we you know we had a workshop yesterday to deploy a whole the data hub data aggregation and machine learning stack and It's basically a one Click deployment of like if we have 13 service stack So you get this Abstraction that's possible as OpenShift and the portability of the full service where you don't have to be an expert on every individual piece and the the interesting piece here is that like so you get these pre-package services, right? So we have the ability for third parties and we we work with many third parties that define applications themselves consisting of multiple services and you can put them up and They are the end customer can then consume them in a way that's Compatible or comparable to how you consume things in the public cloud, but you can take them wherever you want, right? so you you get basically instead of Having documentation how to install things you get the higher aggregation of the full orchestration model down to like pre-encoded Scaling rules and things like that in the cluster as an application you can provide, right? So that that gives us the ability to create the integration and the time to value that you see in the public cloud independent of where you run it And the key here is the ecosystem obviously, right? So you need You need pre-package services Available from many vendors in order to have the same number of service available that you would have in the public cloud so we End up with kind of two ecosystems here one is one is component level It's a download to to build so as a developer I want to download some Java package or an npm and I've write my code and that goes into kind of the application life-cycle management But in many ways in many Scenarios, I just don't I just want to download a predefined service that I just want to operate without having to be an Expert in doing it like I want a database, right? I want I Want storage? I want a message bus. I want Kafka in there, right? It was one of the things we deployed in that service. So So it's it's pre-packaged services that go beyond the traditional binary to the full Spec of the operational parameters of that service and make it easily consumable like you would in the public cloud And and so this gives you it's basically an open alternative to the same model And it gives you a standardized operational model Which you know, so so you know at this point I should have put checkbox in there, right? So so at this point what we solved is we've solved the service abstraction, right? We can package full services makes him consumable very easily We manage the time to value there What we haven't managed yet is the operational excellence, right? And and the standardization of operation that you see in Amazon, right? If you follow the Amazon path, you know you're following an industry best practice and You know, they're like interesting Discussion is like, you know it in open trip. We have a trend to immutable infrastructure for example. So right now you can Right now you can take you can install rel and you can install open shift on top There is a discussion whether that should still be supported going forward or whether you should just have in have a core as immutable host and you can't even lock into that anymore because Usually when you want to log into your host on an open shifting you want to do some things That's not the best practice right because the best practice is encapsulated in what redhead provides you It's it's a debate, but You know in Amazon yet if you go the Amazon way you automatically follow a best practice and We are trying we think there is Kubernetes we can create a standardized operational environment that allows you to then You know Just the same things operate in the same way independent where you run them, right with the operator model We encapsulating operational aspects of an application with the application. So For example, if I'm I'm I'm a database Provider, I ship a database to my customer. I've write an operator for that that operator encapsulates Upgrade procedures and backup procedures for that database right so So that the customer doesn't need to create custom backup rules for example a custom upgrade It's like, you know in windows. There always was a service called HSS VSS volume of shadow. So it's an interesting concept, right? I used to work at a backup company for brief time and on windows You could when you wanted to back up a windows instance running Microsoft SQL Server in a VM You basically had the thing that just said oh, I want to back that up and then VM Where you know in the VM there thing would like go to the service in windows They are you going to be backed up and the windows so VSS would tell the SQL server. Oh, you're going to be backed up The SQL server would flush its buffers pause VMware would take a snapshot of the machine of the data and then say I'm done with the snapshot and The service would restart Right, so you didn't like there was no custom script in that it just was in the system Look at operators as a way to standardize that for services as they do operators to other things But that's one example right the application vendor will set the procedure how to do the backup and that becomes basically a standardized model so that You know if you want to back up your data The system will take care of making sure that it's clean at the point where you can then take the snapshot And so that's what we mean with standardized operation mode. There's another part of that. It's It's I have two more slides so and then we go to questions so So we're standardizing metrics with Prometheus and famous and You know for many things For many things that that you're trying to do in these clusters You want to presently be able to Drive application behavior based on metrics for the full cluster, right? And so that's another so so standardization of how we meter things how we aggregate locks how we Look at for example performance is another aspect of this standardization of the operation environment And then with Kubernetes you already have a standard way to enact action, right? So You know so there are there's an API for everything for every action And so you can automate things on top of these air so we have the way to have a standardized model on how to do actions right gather data and then do actions and so on top of that what Retta is doing we have a service called insights which basically Aggregates information from multiple customers learns from it and then gives guidance to customers based on that So today it primarily Does it for configuration data? So it will tell you whether the configuration you have is bad, right? It's just wrong, you know, it's a or suboptimal configuration or it can oh if it's this configuration It can relate to issues we have seen with other customers So, you know customers with this specific configuration ran into this specific problem and will warn you Proactively about that our last summit we demoed the service where it would look at some of your configuration to say Oh, you're an outlier or you know, 99% of our customers have this configured differently and by the way your performance is off, too Right, and we recommend you change this setting, right? So the idea behind that is that So in a way like if you look what redhead's role traditionally is in open source and you know similar to other open source companies is that we We aggregate knowledge in how to maintain the software provides a stability aggregate knowledge on how to fix the software and So that if you run into a problem if you're a redhead customer you call redhead and you have a high trust level that either We already have seen the problem with another customer so we can address it for you. You know, we may be already Provided a fix if it's a software problem or we are able to identify the problem and then fix this and then So, you know, it's already solved for the next person and basically that's always been the model And now we are taking this model and automating it by aggregating Observations from customers learning from then and providing guidance back So that's the idea behind that is that we can provide some of the operational experience, right? The observational excellence that you see with these big With the big cloud providers or very large ITD partners and we can make that a shared experience for our customer base and grow grow that and so the system will Help you it will have basically a guidance system to help you Achieve the same operational excellence without having to have the expertise always in-house and and that's you know, usually correct performance reliability stability and Correct behavior of the application reliability stability security and performance, right? That's the things you care about and that are all things that we can We can solve to largely agree with this guidance system Now the second piece the next step is self-driving clusters, right? So so putting AI into So into this system, right? So the the first guidance model is possible without the standardization Of course, but it becomes much more effective if you have a standardized operational environment because you So if you if you so insights and traditional rel is Helpful, but of course limited because each deployment is too different So you cannot always make the same conclusions when you go to the model where Where you have an established operational environment with immutable infrastructure, right? Every host is the same every open shift is the same their parameters in different But they are observable and they're in a controlled environment suddenly we can We have a high probability it's something we observe in one customer applies to the other customer as well, right? Or we can measure the differences Now we with a standardized way to measure all of this and the standardized way to act in an act change We can also try to automate that change right the action and enable us to do self-driving cluster So we put learn, you know, we would predict a predictive statistical systems and machine learning models in there to learn from behavior and then automatically react to it And it starts out very simple right with just anomaly detection and things like that So it says oh, you know something unusual happened, right or alert filtering, right? But it progressively will get more and more interesting right from predicting behavior or optimizing Optimizing placement We will be able to for example direct service calls, right? So you can You can If you have we have the standardized monitoring so we can monitor the full stack right up to an application now We can measure the efficiency of an application down to how much CPU heat does it generate, right? So I met someone who wants to do that for machine learning itself, right? So how do you know a machine learning model is good or not? You can of course, you know, they're they're hard factors like how well does it predict, right? How correct is The outcome, but you can also look oh one uses much more energies than the other one So maybe we go with the one that loses uses less energy because that's a factor I care about So I can monitor things old stack or service rooting, you know, I can see machine load and I can push calls For my application to another note because I have You have all the knobs to do that and we can do that intelligently with you know learning from the behavior of the specific environment So that's the ideas behind that So, you know, basically in many ways takes a concept of self-driving car and apply to a much simpler problem self-driving clusters Which will help self-driving cars because you know, it's a data center. So you don't want to suss admin in the trunk, right? Um, okay, it's a cheap joke. Uh, yeah, so so right? So the you know The the storyline, you know summarizes like so Linux's historic role is to be this neutral runtime, right? Open shift does that for the modern cloud world as an extension of Linux The key values that drive cloud adoption are service abstraction time to value and operational operational excellence and the availability of services and we think that we can get To place with a hybrid ecosystem to provide the same And then with the standard as operational model we see there Drive the upper you know provides the operational excellence through Aggregate data aggregation observation and guidance services and putting intelligence in the platform itself And that's possible like because you know because of Linux Kubernetes And then adding machine learning to it any questions Take a very rude But there's a so we joke right like so there's a the head is like so We should change our pricing model and that is a joke right? I mean the city office disclaimer nothing I say can constitute to a product announcement and nothing I say anything like that But but you know someone had the idea why well we should like charge more if you want the root account on your on your hosts Because it immediately means you're doing something. That's not standard. I think it's a cultural change, right? But I think I Really like so I really see the analogy to self-driving cars right which and My you know, I my wife has it has a model Tesla Model 3 that from time to time. I'm allowed to drive And it's really fun So I'm like on the highway in autopilot mode. It has like it's like it's not fully autonomous, but You know, I'm not steering the car and I'm driving well in Massachusetts Let's say 65 on the highway I So 65 miles per hour. I can't you know, it's a hundred kilometers per hour or something something above 100 110 or so and and the car is steering on its own in full traffic so and and and that's that's a behavioral change But You know, it's just the you know the car like I don't have to worry about Anything right especially like in stop and go travel like or that have it like it just Breaks for me and it's much better that than I will ever be it has a lighter and radar Future that look like three cars ahead So it breaks already before I realized that the cars before me are breaking and I think I think once we have These guidance systems in place and and we have automatic Optimization it's gonna be a self-controlling model There's a talk. I think that was yesterday on soft. Was it yesterday? Yeah, so We out of time. So last sentence. So we we are looking at that also for content guys So so this was the operational side, but when you write your own software you you use all these stack components from somewhere and It's the same problem, right? How do you know which version of what will work with which version of what and So we have a guidance model for that as well And we think that it will basically lead people to to the mainstream, right? People will go to the best practice as soon as they have a realistic ability to learn what the best practice is And then people will break out of that only if that's a real good reason to do so All right, I hope this was interesting