Hi guys, my name is Serge and today I'm going to speak boot platform engineering. Before I start, let me tell you how relevant I am to this topic. In my career I went from a developer to technical leader and architect and managed a small team. Then with grow into the engineering management, director of engineering, senior director of engineering and global head of engineering. The size of the teams which I managed and worked with actually grew to more than 100. And together this complexity and tasks and dependencies and other pieces, it's not to show that the titles in your career grow and it's great, but to show the relevance to the topic. I'm still hands on. That's why you will see demos in this particular session with my teams we created different platforms, from mobile platforms to web cloud platforms where actually customers were able to use previously internal development platform. Now they're using this AWS customers running their applications. We'll speak about this later. How to achieve this. And as you see, I'm also a contributor to Terraform, ECMascript and other pieces of software and libraries and frameworks. So let's start. Main question you need to always to ask is why. Always start with why? Why do we have the problem and which problem we have and why platform engineering will solve so problem. Here we see kind of request to Chat GPT what is DevOps? DevOps as you see is methodology. Methodology, yeah, just describe the culture guidance and yada yada. So how do we start? Go figure. Of course I edited here, it was a bit of different answer, but it was long and not exactly prescribed. So I checked the state of DevOps and we have this number of the DevOps tools on cloud native landscape. You can open this particular URL and you see thousands of different tools which you can use for DevOps infrastructure development. The cognitive cloud went crazy as a result. Typically in many organizations we came to the situation when these tools and approaches use it in very very different ways, from different hosting to security to monitoring. It triggered their own problems, right? That architecturally you cannot align and provide the golden path. So technology alignment is hard. Security problem, no way to introduce the security scans and patches. Typically a centralized team. Today security growing and growing and growing in the not investment but attention side. So there is no way to scalably run security checks. Business problem, typical ask from the business is how you can do with less. Now we all have this particular questions more and more we see what happened in an industry. So we need to solve these particular problems and usual problems when it comes to coming to a practice. Yeah, common everyone wants to actually have speed to hire speed to onboarding, right? So the time to the first commit developers want to remove the cognitive load and asking give me some guidance framework inside of which I can just define my front end back end tooling. Give me the guidance. Engineering leaders apologize. Securities went to fix securities problems. If you remember lockflow j or NPM chain problems when some dependency of dependency is actually security vulnerable. And you need to solve this immediately and identify it. Not only solve ops want to do some improvements and business always asking the same can you operate with the same twice faster, delivers the same quality or better and be more cost efficient. Usual story the same. So you see, thanks to experience, I believe on your side it's similar. We can see this from different angles. Okay, so why platform engineering? Why even speaking about this? So if DevOps is a practice, so how can we apply this practice, this methodology, DevOps can be applied as platform engineering. Even Chat GPT will answer this exactly that. You can use platform engineering. So you can try and you will have similar answer. And we finally have data. So just typical question is just like don't speak too much search, show me the data. So the data from the late state of DevOps showing us that the direct impact on development velocity from platform engineering was highly, highly, just like measured and improved, right? Similar. We have about development speed, not velocities, but platform team in general just removes frictions. And to summarize DevOps is sorry, DevOps is the methodology and platform engineering is a so simple examples, right? So ECMAscript is for example a standard, right? And Javascript in the browser or P eight in Chrome browser as you know it is an implementation. So how you practically can do one thing in the way which suits you depending your browser or your ide similar here how you apply this in organization. Okay, so what are the objectives and goal for platform engineering? So platform as a product that's one of the main goals which you need to follow. And first, when you start building platform, make sure that main customers, it's your internal developers and your main metrics are customers net promoter score and adoption. So measure your adoption and how happy your customers what they say about your platform is your metric, right? There is billion ways to measure this. It's a big topic of different discussion. So here is typical steps you need to do to come to platform engineering. You can just follow them. I will not go in exactly. You can download this presentation and check the most important of course to make sure that you come into building cross functional team and you implement your internal development platform. So if you come into this question we need to understand how, right so typical teams topology. If I'm coming to build my cross functional team, what I need to do first after all this time of growth of DevOps adoption, I believe you defined or you will need to define before going to building the platform some prerequisites. What is your CI CD tooling? Is it GitLab, GitHub or other parts? What is your cloud provider? Is it just Azure VMware aws? You define it same your tech stack, your target tech stack. And if you haven't done this make sure that you're doing this before. Otherwise check if you need platform reading at this point and after you can check the team topologies book. It's a great guidance in this particular era how you can define teams and typical suggestion is to concentrate that you have product teams access service or product tests which result some specific goals with clear boundaries. And also you will typically have some kind of organization like that. Very typical example where you have the product teams, you have ops, security and maybe some other aspects which a bit common teams in practice. So when you start with platform Jink, make sure that you find a product which did something what is either already reusable or there is you see some evangelists, people who are eager to help other teams to do similar solutions. If you read the book about Project Phoenix, you will see that some solution which was used by one team sometimes can save all the components. Of course it sounds too just like interesting, but you need to find the product team and parts of ops and security which can define which can bring together in the great platform different aspects of each. What is development experience we need in the company? What is the tech stack, what operations processes we want to have security practices and yada yada yada, put 510 people and start it. If you ask when your engineering organizations have at least roughly 50 people, then start considering this. Of course it all depends on your needs and timing. So at some point you will have platform engineers here. The most important at this point will be your pilot customer. Make sure that you have a product team which will first of all needs your solution and wants your solution. Just like try not never push, just make sure that they do work doing with the pull. Make them the most happy customer ever. Yeah, make sure that they just adopting your solution, they doing the demos, they speaking about you comparing when you shouting about your platform and somebody else. Believe me, ten times ten x better result is when somebody else is doing this, it's more trusted by others. And so imagine you came to first customers, you started being known federal company as a platform. Start working on these directors with products to scale. Make sure that CTO you have as your supporters and then start growing. At some point you will make sure that everyone becomes your customer. Security and Ops are your customer because through platform they will deliver something to whole organizations. And at some point you will start having sub teams inside of your platform in infrastructure, front and back end common services authentication. You will see this will grow. Okay, this is in short what about teams topologies? Okay. And now steps of evolution. So how the hell I start with my platform, right? Typically you have some stages of SDLC, software development, development lifecycle. So from starting point how you do the version control, how you do CI CDCM, all this particular aspect. So in childhood is very first point, make sure that you have a simple developer portal discoverability where everyone can find and discover what is your platform about and what does it solve. And start using your solution and provide some simple tools and libraries how to use your particular solution. It might be something simple, a library, a tool, I don't know a script in the end, but start from adoption, start getting some attention and then you can provide on this portal just like guidance on CI CD testing. Some aspects you can even skip at the beginning. So for example observability or others from my experience for some reason just like left for a bit later to be added into the platform. So they just continue being a bit asylum adolescents. When you grow slowly but surely you will have more adoption, more support from executives and but in from product and other architects. So at this point you need to use something like backstage to provide a scalable developer portal and provide CLI tools to generate your application. CLI UI tools to provide the guidance on your development portal. The limitation team start jumping in. So this is adolescent and maturity. At some point you will see that you need to update and support this developer portal. You need to provide application templates for front end, for back end like react, Java, Python, you define it. You need to integrate this with databases and data lakes, provide the test layers, security checks, previews, environment generations, CD providers, infrastructure as code monitoring on top of all this. And then you will see the most funny one will be proper gradability. Great. So this is the stages to give you some guidance. We will return to actually this maturity because I will give you some guidance how you can achieve some of these aspects. Okay, so as I said, I still want to show you how because at some point it would save me answers. Okay, in GM portal, how the hell you do this particular guy, you can use simple tools like docus hours and let me show you what you can achieve. You can say that based on your particular solution, you created some simple static site generator, right? And you define how docus hours look. Docus hours is a simple static generation. Give me just a second. Which is public free by the way. In this particular presentation I have no affiliates, no links, nobody paid me for this. Mostly free tools which I will show you. It allows you to create documentation with versions. You can say version 12345 of your platform. You see just like it provider you different content in different versions and start from something simple. As I said, you can download my solution, you can download my application simple, just bunch of files or folders. Start simple. Provide more guidance on CI CD. This first step for you. Okay, when you start growing, you need backstage. For the backstage, of course you need something more scalable and typically backstage is also free tool which went open source recently. And here it allows you to provide your microservices, micro applications microfontends. We'll come to this in more details. So this is UI application with database integration with some other aspects. Okay, we'll come back to backstage later. I promise you on front end. So what the hell you do for front end because most of your teams will deliver some front end depends on your business. Of course some of your teams might deliver front end. Yeah, for front end, typical playground is a storybook. So for storybook here is an example. It's also free tool and example of implementation. We for example did this with my teams creating guidewire web framework. Now it's available for Guidewire customers and on preview for everyone. And you start building components. You will deliver pages using this particular components, how to use them, how they have different look and feel, how to use them in for example in the code. All this is easy to be provided via storybooks. Make sure that you deliver. As I said, the way to download your application code and your framework might be NPm packages. You define describe this on development portal and then provide storybook like that. Perfect. And of course you might provide your own templates for react applications view next you decided. So here are some links which you can use to define the different type of applications for your internal customer teams. Great. Then we come in slowly to microservices land which is a bit different beastie. And here we come to an interesting solution which is called Jhipster. I'm not sure many of you just heard about the jhipster, so I will show you this in practice because before backstage it was one of the greatest available solutions. So Jhipster is an open source tool which allow you to also define your application templates and do a lot of integrations with CI, CDCM and other parts. So take a look in jhipster, I will show you how you migrate an application. It asks me do you want to create monolithic application? It might be microservice and micro from end or their combination. So monolithic name is my app one. Let's go Springflux. I will go with more or less default one here is by default it's JavaScript boot application. You can define Python whatever you want. Authentication, you see it's already bundled. SQl sure, go with SQl MySQL. Okay, what about local cache? You see you can do all this particular steps as you wish. Registry. Thank you. No elasticsearch API first here it comes to front end react laundry UI. Sure. And then of course it even just allows you to do different config. Okay, I created this application and I will actually start it. So here in this particular application you see already generated jhipster apps which I did before because it takes a bit of time and it provides developer experience which is super important. So you see when you will do precommit, it already will do the static analysis of your code. That's where you can extend and deliver this as a template of your application and extend this with the best practices you want. Add security here, test execution unit, test execution, whatever you want. Provide the text standards, provide other aspects as you wish. So as you see, it's a simple maven application, sorry, JavaScript boot application is maven. And here we can find my app. Nothing special, some annotations here and here you can also find. And here is web application. Sure, just like typical react with error boundary. You see it's all integrated together, you can do this separately, you can just do this together. Okay, let's see how it actually looks and works. As you remember, we delegated some database, right. And it allows me to provide a user and out of the box I have not only web UI which I can just deliver as an application, but I can manage some users, I have some metrics and even logs which I can deliver to my monitoring system, observability system and just like gather it also second I also have out of the box API which can be used in my developer portal. Great. Why I'm showing this? Because it's fast. Instead of just like describing this documentation to give me the tool. Give me the tool which making this automatically. Right, great. So we delivered all this we allowed imagine just like to create this particular application but this is not enough, right? We want to make sure that we also can deliver these applications. Give me a second, I will show you. You need to integrate these applications, not with different cloud providers, CI CD and yada yada yada. So Jhipster allows you gives you CLI tool which you can also deliver to your customers to integrate with AWS, Azure, Google Cloud platform and many other providers. It allows you to integrate CI CD tooling and provide different integrations with scalp provider. Great. So you can from Jhipster you have ability to use your templates, you have CLI, you have ability to integrate CI CD, great. But there is one more thing in Jhipster and most of the items which you see here can be greatly leveraged in backstage. So we're going back to backstage and in backstage I will stop my server here, not kill my laptop. And here I created my backstage application. Backstage is a simple UI application. This ability to create portfolios, UI plus database plus monitoring. So it was exposed by Spotify to open source built for internal needs and now the fastest standard for platform genius purpose to save you months of your work. So let it start and after that you will see that actually backstage also allows you to create your own application templates. And second, here is backstage application. And you see I took my Myvan application and I put here as a template. I told that you know what, I want to allow this particular template of my backstage which I just showed you. So I'm not bullshitting you. You see my Java application with the same content as before with web app, with index files. Here they are the same files, nothing special, just literally just copy pasted this as a content. It allows you to create your own to have your portal with integrated applications similar to what I showed you before. Perfect. So here we are, I started my own local example of backstage and here I'm going to create my application, based it on the template. This particular app which I showed you before I added as a template so I can create and integrate with my CI CD system. You got it. I am platform developer and I want to deliver for my wonderful teams ability to reuse my application which I created before with the hipster. So let's call it somehow the same, my app name, I will use GitHub, you can use GitHub or other integrations. So here directly you can add this with CI CD. So here make sure that you provide the organization name and you provide your repository. And finally of course it will just start it will try to create the application directly from here based on the template which we created before from jhipster. Great. So first it takes my content of my application. Remember just like I said, java and also UI based on typescript and react. Great. And you see it tries to deliver this to GitHub. Okay, first just like switch the base, it applies the name of my application, maybe some integrations with security aspects like security keys, and finally it publishes to GitHub. Unfortunately sometimes it takes some time. So I will show you that, you know what, I just added a few things. I also added, you know what, I want to deploy my application. I want to deploy my application when it goes to GitHub because it will be pushed to my main branch and I want to deploy this to AWS, I want to provide to publish my web to three s, three bucket and then expose with CDN. And also maybe I want to do something else. But we will see. First of all you have a problem, maybe something happened. Let me check if something what went wrong? We learned something, right Kush? Why? Anyway, I will just delete my application, so you will see that I'm not bullshitting you, just try. No worries. Who knows, maybe there's time it will work, maybe not, you'll never know. So I added the deployment to AWS and I added terraform to my template. Terraform is also just can be your kubernetes, can be separately, can be for your kubernetes, some particular databases or other. So here I also just like it because it's web application with some database. I added the publishing to CDA to cloud from and I want to see all this in action and I hope this time it will push the directory properly. Things happen in this life and we always can learn something, right? Okay, so once again I created a template with jhipster. I use this template in backstage, I added integration with CI CD in the way which I wanted. With GitHub you can use GitLab. I use terraform for infrastructure as code. Perfect. So here is repository and let's see what happens this time it just like pushed off. So you see I'm not bullshitting you, it just directly committed to my application right now. And because I added this wonderful GitHub flows what happens. My CI CD pipeline is code started to give me fruits and just install independency and then it build and deploy to AWS. Perfect. It will also use terraform of my choice. Great. Of course I just did this before because it takes a bit of time and here it is you see this particular URL is already published on Amazon cloud front because it was used in my code and directly said that where it should be deployed on AWS. Here you provide secret keys and you provide the s three bracket and distribution perfectly. So from local host to actually published application. Done. Great. So after we got all the steps, what's next? It's upgradability believe me or no guys. But this is 1 million question and I mean it. I mean it because literally when you have small platform, your yearly cost of the updates for your internal only customers is measured as up to 20,000 dependent. So when you have medium size you already have $80,000 per year, which might be cost of a salary for a few people depending on the location. So when it's large organization starts growing for up to half a million plus dollars per year, which is crazy amount of money. Question how you leverage this, how you may invest some research to decrease this cost or eliminate this. And of course it's not only cost, it's development frictions. Yeah, just like if you want to make this smooth and deliver your updates for customers, you need to make sure that this is as smooth as possible. So what the hell do we do? That's where backstage is not exactly the great solution. So of course it added our portal, you see just it provided documentation path. But when I'm publishing a new version of my application, maybe I changed some API, maybe I changed some imports I want to update. So how the hell I do it go do go here. That's why I mentioned for you Jhipster. Jhipster allows you to say you know what, between previous version and this version there was some changes. So create a branch of application developer code base. So imagine product one code base try to apply my changes updates, I don't know imports API and provide them the branch with this updated version so then they can figure out what to do next. Okay, this is one part, but sometimes changes are very specific. It requires not only git changes but maybe changes in the code to not follow documentation precisely. For this you always can use code shift. So code shift allows you to say okay in the language of my choice, I want to do some code transformation transformation just like on JavaScript transformation on Java on your needs. So they procreate some ats three from your files and tries to apply the changes you describe here. For example, react code mode is a JavaScript react implementation of code shift. How you can do this? For Javascript there is Java implementations and others. And for dependencies management you can use renovate it's also a free tool which support crazy amount of languages from Python to Java to node js and anything. This is just great. Which allows you to say okay for different reasons, you want to update your dependency or dependency of dependency or you don't. Great. So all these particular pieces giving you a strong platform. But you know what, as I said, it's just like sneak peek pick of all the pieces I created for you when I was preparing for this particular presentation. A repository, this created list of these tools and blogs and actually different informations and links related to platform engineering. All these pieces, how you provide code templates, how you create microservices front end, how you provide upgradability and apply different aspects like I don't know, Kubernetes provider, some infrastructure platforms, how you do all this here. So make sure that you go in here and just find the best tools which you want to create your platform. If I had this at some point it would save me many months. But now it's time for me to share my experience. Great. And from this great experience, sometimes that's how we call our mistakes or learnings. Let me give you a few advices. So make sure that pilot team you are super involved with them and they become your best friend and evangelist. You will get a customer and from customer you will get evangelist at the top of your customer ladder. Treat your platform as a product as soon as possible. You need to deliver business value business. I'm sorry, between us very honestly, not very often cares about technology. They care about the product, about the solution, about customer marketing, sales, technology. It's your responsibility. Make sure that you solving the business problem. So business problem with adoption is speeding up. It's just like cost all those goodness. So for this, treat internal teams as customers as soon as possible. Provide them slack, provide them the developer portal, invest in platform PM. Just very great topics, very big pain to be honest. Grow either internally PM or find the guy who is crazy interested in all these pieces and know how to speak with customers and how to build platforms or ie is eager to do this evangelize. So you need to send regular releases, you need to be predictable. Do the demos newsletters, make sure that others speaking about you as in any other product, it requires the stages of evolution and few interesting things. So for UI adoption would make sure that you put teams with the UX. So for example, my best partner in crime was head of UX. Make sure that your developers with UX working very closely that they give the design. Then you have front end components for this same for your ops team. Yeah. Just so if Ops teams is saying we start implementing some monitoring solution new relic, I don't know. Just like Datadog, make sure that you provider in the tools for this in your platform and this goes together. They're speaking about you, you saying that you're enabling them dumb down just like win win and cooperate with all the architects, directors, VPC here. Actually you need to make sure that from the top down you have ability to enable bottom up. That's the thing you are looking for. And jackpot strategy. This is my favorite. Yeah. If you find the way how to deliver the platform which previously was internal to your customer, you not just cost center. And where companies see the spend, businesses spend, they see the profit and profit will be just like your friend. So make sure that you find in the way how you can deliver your goodness to customers. This way you will have test based inside of your company great fast feedback and you will have customers also win win situations situation for everyone. Okay, so to summarize, remember this maturity of our platform. I gave you some ideas how you can provide this portal solution for discoverability. How people can find your platform and start with it. How you can use your application template, how you can integrate with data and provide test checks. Because you can include this in your CI CD pipe as code. You can add CD and infrastructure as code inside of your flat templates and in backstage or jhipster just added separate steps, template creator creation and CI CD of your choice. Code observability also should come there. Tracing should be integrated. Everything should be just going to backend and front end and infrastructure immediately and upgradability. You saw jhipster code modes and other solutions can help you with this. As a result you will provide technology enablement so it's centralized way to introduce new stacks and libraries. You want to migrate from Python to Java, provide a new type of application and start actually working on the feedback. New cloud adoption, security and compliance perfect example. Look for J or NPM versions problem we had before with the security you publish update in the platform in the centralized way and immediately through this upgradability tooling. Q. Deliver this to all your customers internal and external. Vinvin perfect. Amazing. And also DevOps will be happy because they can introduce the new layers together with security common security checks across all your organization cost one you need to measure this but it's just literally we have state of DevOps, we have research, it reduces cost of onboarding technology alignment pivoting. If your company acquired other company and want to integrate or want to change the tech stack and upgradability cost is just a clear win result. Okay guys, what I can say here are the links to the slides which I just showed you together with the resources which I created as a GitHub repository. The best day to start is always today. Make sure that you start in your journey. It will really take thousand miles, but you will get to your mountain, to the top of to your dream, right? And make sure that you're doing this first single step. I gave you some guidance, shared my experience. Good luck. Now do it. Build it. Have a nice one. Bye guys.