 Hi everyone, I'm Gordon Haff and I work in Emerging Technology at Red Hat and I'm William Henry and I'm working in the Ecosystems Engineering at Red Hat. So what we're going to take you through today is several relatively newish initiatives, concepts, sort of sets of technologies and practices that we're putting together that are sort of collectively around making it easier to consume open-source software. And this is our agenda. I'm going to set things up a little bit and I'm going to talk about those initiatives and then going to leave you with a few kind of takeaways about if this peeps your interest things you might like to try out. So to quote somebody I probably quote way too often Stephen O'Grady back in 2012 he wrote this sub blog post the base he said do not underestimate the power of convenience and one of Stephen's arguments which I think I disagreed with at the time but as usual Stephen is right and I'm wrong was he was arguing that Napster for example wasn't so much about getting free music but was because he didn't have to go to record store and I think the popularity of streaming and I thought he was wrong but the popularity of streaming I think argues he was right now he was arguing here really about open-source versus proprietary softwares you won't try out open-source he downloaded and it's easy peasy versus you have to call a sales rep and between us no one likes to call sales reps but the interesting thing is though is I think you carry this forward and the same kind of logic applies to installing and operating open-source software versus running a cloud service of some sort of cloud native service the software is a service and in fact Stephen recently very recently wrote another blog post basically about integration versus best of breed and I think that plays in here as well and I think there's a sort of central tension here and I will describe his attention I actually have a graphic I really like here with the cloud punching out with this hippie looking guy as the cloud versus open source but I think that actually overstates it I don't think in general we should look in this as a conflict but rather different ways and a continuum of doing things so rum or you want you operate it you can freely modify it you have you know how many desktop managers are there for Linux it's portable you can run in different places it's open source but it tends to be less convenient there's more work involved here in for instance freely modifying whereas at the other end running a cloud the cloud takes care of all the services for you but you have to you know this is the service that Amazon offers no they're not going to tweak it for you you're going to run what they have they decide what types of services you should have or what you really need they often are specific to a particular cloud so while there are commonalities particularly if you talk about just compute instances among clouds you start talking to AWS or Google or Azure specific services and they're fairly specific and although they may be built an open source the services themselves are often not open source and you know certainly simplify this you know even more you'll get you'll just freedom to tinker with your software and you can kind of do anything you want it but sometimes people just want to get in the car and go places they will use the car as essentially a utility at the end of the day so I'm going to start first talking about something called operate first Williams going to then come in and describe some solution blueprints which is a project that he has been very involved with and then I'm going to close this out talking about cloud services now the basic concept here with operate first is you know before open source you know the code had all this proprietary value associated with it the the value of the software was in many cases served in its proprietoriness that there was this artificial scarcity effectively created by licensed software that you know yes it's essentially free to reproduce but you have to pay for it anyway and I think arguing arguably open source level the scale and I should probably interject at this point and Sam borrowed these slides an example of open source collaboration from my colleague Carsten Wade who is actually very involved in setting up by the community some some of the community related things related to operate first this is still fairly early days by I'm going to show you some concrete things that we've achieved so far with it but this is definitely going to be an area to keep your eyes tuned on and the idea here is you you know it's you leveled the scale because if everyone is access to the code having access to the code isn't the differentiator it's what you do with that access and what we've seen now in sort of the cloud computing market and I'm using this term broadly we could be talking about hyperscalers we could be talking about software as a service we could be talking about other types of outsourcing operations including more traditional manic services but it's the operation of the code that becomes more valuable if you know anybody can you know install some you know kind of name your database name your AI ML type of service you know anybody can do that but what gives Amazon or a or Amazon or Google or Microsoft or other or you know arguably even someone like Salesforce really what gives them their competitive advantage is that they can operate that stuff at scale really really well and certainly better than many particular particularly smaller to medium enterprises can do so so if this ops is tilting things in favor of a cloud provider what do we do to maybe let you know level things back again and this brings the stop rate first and we were actually talking last night we're still working on the language of a law of this stuff so when I say to operate versus a concept that doesn't mean we haven't actually done anything real but we're we're trying to think of it as sort of what we're conceptually trying to accomplish at this point and Karsten who just walked in is looking lost is our community architect who is working in this so he's he's going to keep me so if he starts like waving his hands and like this I'll maybe ask him to say something but the idea here is ideally you know we've been we create software and certainly we all have quality assurance and various other types of testing but historically this it's been let's find the bugs it hasn't and it tends to happen at a level that is a long way down from what some of our bigger customers are doing so the idea here would operate first is let's try and think about putting operational experience and really operational excellence into software development for the start and that's by extending development to include all this other operational stuff in a production environment and production is another one of those words that's kind of a bit funny and we've argued a little bit about what that means but I think again conceptually I think we can think of it as it means there's lots of stuff connected together in complicated ways to the degree that you can kind of bring out unpredictable actions potentially in software and it works by bringing code to a production cloud they're working with the cloud operators the SREs basically of that production cloud and they experience firsthand the operational consideration so you know certainly DevOps DevSecOps is one step towards kind of bringing some of that operational knowledge back you know kind of back into a developer but this is really the idea of doing this at production so that developers can start thinking really very far back in the process over oh these are some of the scale considerations I've never been exposed to that directly and you know if you're presumably a developer at Google or somewhere like that that may be sort of natural knowledge or firsthand knowledge but often isn't with more traditional software development now I'm not going to even try and explain all the organizations involved here we basically I'm from Massachusetts and basically we've been very involved with something called mass open cloud which is it's a consortium of universities and other kind of research institutions in order to kind of set up these again production level research environments and it's tied in with the open infrastructure foundation which is actually you know hosting some of this stuff there's the Massachusetts Green High Performance Computing Center out in Holio red has been involved in this we're also directly involved with a number of participating universities Boston University we really kind of kicked off this effort and Tufts is another one we've been working on and telemetry so there's a bunch of intersecting people here which I won't attempt to decode but the sort of the message here is this is not kind of a red hat thing that we're doing by ourselves it's something that we've got a lot different both commercial Intel's in there IBM has also donated some power hardware I believe to some of these efforts they're all kind of working together and this really kind of widely recognized problem you know as I said this isn't just a concept at this point so at this zero cluster for operate first it reinstall ran ill some issues they were able to debug with the ops team and the projects that were operating first they resolved this problem within a day and there were some there were some failures including with the Kubernetes Advanced Cluster Manager and got some workarounds put in place and obviously Fed changes to need to be made back into the upstream because that's how we do development and the ACM team basically noted that if this had happened at a customer site after the software had already gone out the door this was going to be like a total bear to fix because we wouldn't have been easily able to replicate the customer site so this is the kind of thing we're hoping to solve with operate first so with that that's how operate first this is a fairly early stage project and as I say Karsten is a community manager and we will have some we will have some links later on but if you see something here that sounds really interesting make sure you grab Karsten and talk to him so with that William off to you thank you Gordon so one of things that we've been working on internally within redhead engineering is to try to figure out how to make this problem sort of easier for customers right just some terminology here Gordon mentioned blueprints solution architectures is up on the slide there is a lot of terminology going on around this problem I'll try to explain it in in terms that makes sense but within red hat everybody and their cousin wants to have a solution architecture or a blueprint everybody knows what has different interpretations of what that is we're going to be announcing an initiative in a couple of weeks at kubecon which will have a name around it that might have the term patterns in there but these are all from my perspective these are what you're going to hear today these are all interchangeable at the moment but patterns is what we're looking at so what we've been trying to figure out is these architectures are really really hard right we know this from from a long time ago but the cloud native environment as well right we're all moving towards sort of cloud native dev ops get ops all that good stuff right the landscape you often see the the cloud native landscape picture with all the different projects in there but even without that if you look at a specific deployment of an advanced architecture for an enterprise system it's very very very complex it's huge how many people here have experienced that okay by the way for those of you on the stream thousands of hands went up right there thank you everyone okay so actually one but that was good anyway the point is is that when I look at it like you know they call me a senior distinguished engineer or whatever when I look at this stuff it's really hard there's a lot of layers Gordon talked about ACM their advanced cluster management we talk about things like Argo CD what we call OpenShift GitOps we talk about pipelines things like Tecton we're talking about Open Data Hub which allows you to do data scientists to do machine learning in the background which again is a very on its own a complex project and then all of the other pieces like messaging and data lakes and making sure you have connections out into GitOps and image registries they're very very complex problems right and we don't want people to try to reinvent this all over again if we're going to try to solve this operate first problem we want to be able to try and at least deploy things in a very predictable and automated way we see these patterns emergent right so when we're out there with customers we go this is a very interesting pattern it's something that maybe we should be able to take inside at Red Hat and try to replicate and see if we can turn around and first of all tell our consultants and others in the field hey this work that customer XYZ go and try to implement this at other customers as well learn from what they've done okay so one of the how do we do that well we find these successfully deployed complex solutions that we've we've discovered at a customer we do not want William Henry in engineering to be dreaming this stuff up we went down this road many times at Red Hat with what we call reference architectures right where an engineer goes if I were deploying this this is how I would deploy it turns out perhaps that it might be a clever way of doing it but customers are usually right and they will inevitably deploy things a different way than what we would do so let's take something that's out there and that's not just a single product it's multiple products and let's document the architecture let's review it with a lot of peers internally engineers what we call communities of practice a lot of our solution architects senior solution architects and consultants in the field and then we want to turn around and take the code that they've developed and not and this isn't just application code this is all of the configuration code how to tie things together all the lovely cloud native yaml that you all love and move that into a framework that is repeatable for doing DevOps and GitOps and then we want to publish it as open source and say hey if you want a data center to do AI ML here's the entire blueprint pattern whatever you want to call it and it's more than that though we're going to actually do a little bit more than that what we're going to do so on the left-hand side here you're seeing that whole kind of front-end gathering information piece validating piece publishing it out to the world not the code to say hey this is a wonderful architecture you might want to look at it but of some of those architectures that we want to invest in because we see a lot of them in the field very repeatable we want to take some of those and automate them move them into a Git repository where we're continuously testing them so one of the downstreams from this patterns from these patterns are quality engineering so we're going to be at Red Hat running QE on these patterns continually to make sure that an upgrade for example in Kafka isn't going to break multiple manufacturing sites around the world right we want to make sure that that can happen we want to be able to look at these things in a much more holistic approach rather than single product problems or single product failures what happens to this architecture as products upgrade as things change the other part of it is going to be we have at Red Hat a system called or HPDS which is our sort of demo environment for people labs if you go to Red Hat if you go to summit or other places that do Red Hat labs it runs on that environment as well so you'll have you know 50 people or 20 people in a class all logged in doing examples of code and working on it that's all in or HPDS and of course the other downstream then is the community itself people who are who we want to get involved in these particular patterns contributing code to them updating them to us maybe there's a pattern that has a a specific open source project but a third-party partner says hey we want to actually show you what that pattern would look like with our open source project in there maybe it's a security project for doing examination of images and looking at registries and see if there's any failures or problems within certain images but also customers or partners we want people in the consulting world to be able to turn around and go well gosh why would we reinvent this now not only are there open source projects products but there are also these these patterns that Red Hat and the community are developing that we can turn around and take and deploy and modify as fits our customer right so they're like going to be great starting points here's the type of one we were working on I don't want you get too much into details of this but you can see this is very complex we're talking about on the on the this is a what we call an edge use case I'll talk to the slide I don't want to get too much detail about it but you can look it up later is this there it is on the data center side so we have a data center let's say this data center is a factory sorry is a large sort of main factory in Germany for a car manufacturer something like that but they have other factories all over Germany or other parts of the world and they need to turn around and do anomaly detection on various machinery that's going on on the factory floor right and so on the on the the problem you have here is on the data center side you want to have data science to be able to do machine learning so you have data scientists using things like open data hub etc you're gonna but you're gonna have to have pipelines for them to do DevOps right so that's things like tecton you're going to want to get data back from the factories to do more machine learning later so you're gonna have distributed streaming services you're also going to want image registries for pushing things out source code repositories you're gonna want to get ops controller to watch what's happening on git and trigger events to start off both you know image building etc and then you're going to want to have smart factory management in other word for advanced cluster management where you're looking at how many factories are coming on how how do we know when a particular factory comes up what to deploy on that particular fact and of course on the factory side you're going to have your applications various messaging and integration pieces to your machinery in the factory the what they call the line servers and stuff like that you want to share storage the green aspects there are kind of like funny enough the user code not an awful lot of user code here an awful lot of other pieces things like messaging brokers and again distributed streaming and MQTT etc very very complex right now watch what happens when you actually try to deploy this that's a logical diagram this is just one view of that particular thing from a this is the physical architecture for the DevOps piece I believe but there's also a GitOps part of this too so when this is for when the data scientists do their updates to this and you as you can see various different networks we want to segment off different pieces of this for security etc so different networks different cardinality of things that are supposed to be highly available and of course you have the message flows that we're detailing between them so what we try to do is document the architectures of these diagrams so we all have the same semantics when we're talking to each other about these deployments of what they look like in the patterns but that's and that's more of the hey here's an architecture that works go out and talk to more customers about it then we take the code that's involved in that and we want to bring that in and automate it and let people use it and try to make it as simple as possible to deploy we have it down now to like a update a file with some of your details your keys and stuff like that and then do I'll show you in a moment and then do a make build but we're even going to modify it more the idea then as we get happy architects right they go out what I would like to see as architects coming into something like the OpenShift user interface and saying okay you know I'm looking at my environments I'm looking at my current architecture what is what is red hat a half for these patterns let me have a look oh here's a manufacturing pattern that's doing a IML wonder what that would look like and being able to turn around and essentially push a button and deploy that out into an OpenShift cluster really really easily right it would just do it for you right then you can turn around and look at it and say okay now that application is anomaly detection for a for a manufacturing business what if we wanted to do this for a health facility where we're doing detection of pneumonia well we're doing an example around that as well we'll be codifying that in the next month month or so an example that was happened at the VA and they were doing pneumonia detection and it turns out the data center looks kind of the same but what happens out on the medical facility is different to what's going on in the factory obviously we're talking about not just data coming from machines but we're talking about images coming from x-ray machines right so there's a lot of file-based stuff going on there but then you turn around you can so you can tailor it to make you mean oh okay we're not doing pneumonia detection but we're doing we're doing let's say TB detection okay so it's going to be a slightly different algorithms or whatever but mostly it's the same thing oh we're not going to use S3 as a storage we're going to use something else but you can tailor it slightly to what you want and then you can push that back into your GitOps environment and now you have a working by the way when we talk about deployment we're not just talking about deployment once we have full day-to-day operations where you are able to turn around and modify the code and it will continually update right so how do we do that you'll go out to the repository and you'll find this pattern that you want out in the pattern repo you can fork that yourself and say okay this is my particular fork of this my company's fork of this and then you know as a developer or an architect you do a local clone you can go in and change a few secrets in there like hey here's my github information here's my quay our key registry information other other secret stuff like that obviously you're not going to push that back up to your repo that's not stays on your on your laptop and then you can simply press you know run make deploy or press a button and off it goes and what happens is it will take some very very complex pieces that you as a developer are an application provider don't care about it's like I don't really I'm not in the business and this gets back to the operate first stuff that Gordon was talking about I'm not in the business of managing all this stuff get ops and pipelines and messaging brokers and all that let this do all of that I do want visibility into I need to know what's going on in there my developers want to know but at least get this thing up and going what what's going to drive that open shift get ops which is Argo CD project and it's going to turn around and deploy different applications which turns around and brings up different operators like the AMQ operator are the in this case even the advanced cluster management operator so we're driving advanced cluster management from our open shift get up so that we're using open shift get ups as a consistent way of deploying everything out onto a cluster and then of course all the other pieces of applications like oh we need the pipelines any the data lake set up I need all these other pieces so one way of looking at a lot of people when you go into something like a Kubernetes environment you start looking at pods and things like that right are in open shift you might look at the operators and all that water installed etc but when you're an application developer you're really kind of looking at oh I thought we had hidden that but there you go this is actually what happens in the same environment when you have that edge whether it's the medical facility or whether it's the factory you can simply turn around and import the cluster in and what that does is advanced cluster manager advanced cluster management turns around and says oh okay you you want to be registered here as a factory great it installs an ACM agent out there the ACM agent says oh what one of my first things to do is I'm going to get you up on with open shift get ups Argo CD and as soon as that happens it says hey what what is this where am I oh I'm in a factory oh if I'm in a factory these are the applications I download so downloads the applications which are to do with the anomaly detection the notification service those things you saw earlier but part of it too will be turning around and saying things like hold on get ups applications it will also be installing certain operators too for things like MQTT camel K and open shift our site MQ streams which are all part of that manufacturer say so literally three things that you would have to do in this edge use case one is change your secrets deploy the data center and register your edge cluster and you're done it deploys the whole thing for you it's pretty cool stuff I have a question here are we doing questions now sure yeah let's do a yes absolutely repeat the question oh sorry you're asking at any time can I query and see what am I running is that correct yeah there are a there are a couple of different ways of doing that one is you can look at it from inside of the open shift environment right so you can turn around and see what's actually running in open shift right now what operators running what pods are in there and you can there's lots of different ways within open shift or Kubernetes to say hey show me the manifest of everything that's running right now and all that good stuff the other way of looking at it is from within something again this is a picture you probably can't see down there but within Argo CD it's a different way of looking at saying hey let me see this from an application perspective right now when you drill down so for example there's the Manila ACM with the Manila is the project that we call this so Manila ACM literally all that's going on and there is ACM is advanced cluster manager up however here in Manila test all application that has a whole bunch of things about staging and development to do also to test and within that you can see a complete graph of everything that is working or not working or synced or not synced within your environment so it depends on how you're coming into it you can sometimes you go in here first to see is everything running yes now let me go into open shift and see what the manifest is and what's run what's happening oh absolutely yeah you could absolutely do that without a doubt and you could go into the you know again you're going everything here is in Git everything the code and all of the operators all the YAML everything is in Git so you can go in there and modify that and once you do a push to get it will turn around and Argo CD you'll see the change you go oh what does this mean I need to synchronize things yeah so I mean the question is sort of you know software bill of materials point I think what is fair to say is all the tools are there to do it they are probably not as well integrated at this point you know where it's like you know press the easy button and get your software bill of materials but those are the kind of things that are being worked out yeah yeah well at the same time though you can turn around and say hey within this namespace on my cluster tell me everything that's in it get a full manifest that's like in fact that's how we help debug this stuff we turn around always and say okay let's run in right now and we go through it and go oh that's why that's what's causing our issue so yes we're already doing that today but in fairness to your point one of the things we're discovering is for troubleshooting this ourselves and walking through that sort of tree of finding where things are we're gonna have to do a better job of documenting that right now though the manifest these visual tools etc are the things that we we find the most useful part of it so again within I think we're accidentally going through our hidden slide our hidden slide so again this is the sort of open chips view of that yeah so just in summary where you can find information on this today again this is very very new but oh hybrid cloud patterns dot IO and that's with that's hyphenated hybrid hyphen cloud hyphen patterns dot IO and github up there there's the hybrid cloud patterns link as well the project there and we've segmented it out into like common areas documentation and also then these particular patterns for manufacturing or for we're going to be doing the medical diagnosis we might be doing some smart cities examples as well and so the idea is you can go in there see the entire most of these cases involve multiple data's multiple clusters so we're talking like data center plus a lot of edge cluster as well and so you'll be able to just go in and see these things and of course they're also consuming from a kind of a common area as well so we don't have to like rewrite the the yaml for all of these AMQ or MQ streams as well any more questions on this before I go on because it's but it gets back to what what Gordon was saying earlier the operate first kind of example where how we do in time good yeah yeah so it gets back to his operate first issue what he was talking about earlier is we want to try to make deploying these very complex things simple and also have a pattern that's recognizable and reusable over and over again so that even if yours if the complexity is different from one deployment to another how to get it in there and get it deployed in the consistent manner and a supportable manner is is simple or at least intuitive yeah so sort of fish this off let's talk about cloud services so this is all been great these are ways to simplify complex deployments but you're going to have some organizations to say I don't want to hire a bunch of SREs you know I don't want to operate a lot of this stuff now it's not that cloud services and cloud native services and public clouds are necessarily simple certainly if a buy is looked at the AWS services page recently with however many forms of Kubernetes and container operations there are there it's not exactly simple but still the underlying operation of the infrastructure is something that a you know an enterprise organization doesn't need to do and so there are a few different scenarios here so you know I guess we'll use it if it breaks I want to be someone else's problem I don't want to have my engineers being woken up with a page at 2 in the morning because some underlying services broken up of course your application can still break and things can break at that level but still the infrastructure takes a lot out I you know you don't want to talk to a sales rep I you know also I don't want to talk to a sales rep in order to use it I just want to be able to open a ticket to use there I just want to have my credit card in a in a service and do some clicking on a self-service console I don't want to have to worry about have the latest patches been installed in the underlying you know operating system or other infrastructure I don't need to worry about worry about upgrading it but here's the critical distinction so so far everything I've said is pretty much the case of if I went to you know name your favorite hyperscaler or other large cloud provider the differences but what if I want to maintain a choice of cloud providers and an option if I choose to accept it down the road to maybe operate this myself maybe some portion of my application for whatever regulatory reasons I really will keep in house but I'm fine with going out to a public cloud for kind of the more pedestrian type add one thing to sure so this is really important too because if you look at what I just talked about we're talking about these very very complex architectures but I did mention that a lot of times this is installing things like AMQ operators or AMQ streams operators or the open data hub operator and all that good stuff and that that's all good and well but we're also looking for from those patterns perspective in the future how are people going to be using this and do I want to turn around and swap out the AMQ operator i.e. installing it and doing what Gordon says here and says I don't want to install it I don't even want it to be part I want it to be mentioned in my pattern but I don't want the pattern actually deploying it well then it can be switched over to a managed service right or a cloud service where we say hey just get the AMQ that's up in the cloud versus the one that I'm having to install and manage as an operator so that's another another aspect of where we're looking with the the patterns that I talked about for perhaps next year you'll you'll see that stuff yeah and this is the basic architecture and don't worry about don't pay too much attention all the various logos here but someone already made a nice diagram of this for me so I'm borrowing it so you know Linux underlies a lot of this stuff Kubernetes is the standard container orchestration layer on top of that you then have what we call you know various cluster services associated with with Kubernetes and then on top of that and we're just starting to build this out right now but various types of services so platform services this and this is how we've broken it down platform services application services data services and developer services and I'm going to dig in a little bit deeper into the data services because it shows kind of how operate first and solution architectures and cloud services all kind of loosely interrelate with each other and then top of this you have multi-cluster management you know what the ideas here is and what it frankly challenges as we put together these cloud services is that frankly one of problems with a lot of cloud native services is they're not very streamlined from a usage perspective and and the reason is partly organizational and just how number the cloud developers have developed their services namely you have your two pizza team or whatever developing this service a different two pizza team developing that service and it actually leads to a lot of innovation because you have these teams that are kind of left at some level to do their own thing as long as they meet whatever their objectives are but it doesn't necessarily encourage working with other teams to have an overall streamlined experience what it one of our goals here is to have a much more streamlined developer experience across different kind of different types of cloud services here so in the case of data specifically the way we kind of think of it is that you get data rest data in motion and then data that you're actually working with that you're doing some something with and so those are kind of the specialized data services that areas that we are thinking about and I'll specifically talk about primarily the data work somewhat data pipelines and then you have underlying infrastructure under there so for example you're probably most notably large pools of typically in our case set based storage that you're kind of using to load all this data that you need for AI ML modeling and training and so forth so let me give you an example here so open data hub is open source it's essentially a curated although I know with size of that diagram it doesn't look very curated but there's actually an amazing number of projects happening in this space in terms of AI and ML in terms of training in terms of data analysis in terms of metadata management in terms of the underlying storage in terms of basically data in motion streaming and so forth security and governance kind of overlay on that monitoring and orchestration overlay on that so you update the open data hub which is essentially this collects of upstream projects and you know we do work in the upstream in order to integrate that now this looks complex yes it does look kind of complex and I think our serve their original plan was what we typically have historically done at Red Hat which is well we will come out with a product we will productize this open source project and provide support and provide knowledge and provide blueprints and do all that kind of thing with it and I'm not saying that is not going to happen at some point but it right now is an open source project we stand we decide to deliver as a service to start with so basically upstream code and it is the upstream code this isn't you know there isn't proprietary stuff in here but is enhanced with operational excellence so you've got the open data hub upstream we have a subset of open data hub operate at scale for the community and university audiences and that's how we're going to get to that operational excellence to infuse into it as William mentioned us some of the solution architectures are you know are kind of going off in the oh you all operate it on Prem the open source project on Prem here are some things that we'll that we're doing to make it easier to do that but we assume that you want a curated open data hub that you don't want to have SREs that you don't want to have your data engineers necessarily operating a large-scale infrastructure to run that you just want to you just want it to run and over time be able to run it in multiple places right now this is an Amazon web services which is where we've historically kind of tended to roll out things first just because AWS was kind of furthest along we started rolling out various types of services and various types of offerings and clouds and so we end up with a cloud service offering called Red Hat OpenShift Data Science that's delivered as a cloud service on our managed Red Hat OpenShift instances which in turn run on AWS where we're not a data-serre operator there are companies out there that do they have a lot more expertise in this kind of thing than we do and this is I think that's nice because it gives you a kind of a sense of how you can take things in various directions while not kind of having to make a radical break in order to use oh well you have to run this for all time on this particular cloud service or you need to use something else or you have to use an open-source project you have to figure out how to install and operate on your own and I think this guy shows this idea of choice among all these different footprints yeah it's a good it's a great point too because again it goes back to if you adopt one of these patterns that we were talking about these solution architectures that use open that use open data hub you can turn around and say okay instead of installing it let me just run it as a as a service yeah and for that matter you would be my very well with people who like hey I won't give this thing before I go a lot of trouble let me give this thing a try and okay that's nice but I want to substitute out these other open-source projects and you know I go to trouble of saying up my own installation so that is it these are some links William already gave you the links for solution architectures we do have an operate first cloud website which Karsten here is frantically expanding and making more community friendly but it's a good place to get started I'll also give a plug of in the associated open info labs these are under open info foundation there's a telemetry working group which is obviously very important for all these large complex distributed computing things so again that's not as well-developed as we'd like it to be either but if telemetry and distributed tracing and so forth the stuff to interest you by all means go check that out as well so with that one more oh oh yes don't forget to go down you can get Gordon's book there's a few copies of my open-source book left that will be our booth tomorrow and there's also tons of we have the best stickers you have the best lot if you don't have stickers you got to come to the to the booth and they're also giving away a wireless charging mouse pad whatever that is okay and hats and hats lots of hats so please yeah come by the booth if you have any more questions so thank you all yeah thank you we're going both be heading over the sponsor showcase from here so they will probably be meandering around Red Hat booth and so forth so they buy his any questions feel free to swing by thank you very much