PurePerformance - Recap KubeCon 2023 NA, State of Platform Engineering and more with Andi Grabner

Episode Date: December 4, 2023

Besides attending KubeCon 2023 NA Andreas (Andi) Grabner, co-host of PurePerformance but guest today, has also travelled parts of the US to chat with the broader observablity community on topics such as Platform Engineering, Observability, DevOps, Automation & Security.Tune in and get a quick recap of all the topics Andi has picked up on his recent trip

It's time for Pure Performance! Get your stopwatches ready, it's time for Pure Performance with Andy Grab to another episode of Pure Performance. My name is Brian Wilson. And as always, or as sometimes I should say today, as sometimes I have with me today's guest, Andy Grabner. Andy, as guest, how are you doing today? I'm very good. I'm just trying to figure out if as a guest I have

to play a different role in here now or do I also need to deal with your bad jokes or strange dreams or do you then, I don't know, introduce, no, you just introduced me. I don't know. Yeah, well, this has been fantastic so far. I can tell you that as a guest you're failing terribly. Most of our guests do a lot better in the beginning than that. So, I mean, you have been a guest before. And what people usually know about when you are a guest, it's because you've done some traveling and seen and heard some awesome things in the real world. So recently you were just at KubeCon, right? You were at some, one of our new favorite topics, this platform engineering. Was that a convention you were at? No, I was doing a workshop tour and I'm actually still in my tour as of the recording. So I did four cities on the West Coast. Actually, I'll make it all the way to Denver, right? I was supposed to meet you in Denver, but then unfortunately you couldn't do it.

But yeah, so a lot of platform engineering discussions we had lately with our community. Thank you, Kahn. And next week, I'm heading to Texas

for an additional continuation of the topic. And then the week after, I finish my trip in Hungary, Budapest. Having, again, platform engineering

as a topic, as a keynote topic at a financial conference. So, like, conference for the financial sector talking about platform engineering.

And then you get to go home? I think so. Yeah, no, it's all good. You know, I'm doing this because I like it. Seriously, it's really good. You know, I'm doing this because I like it. And seriously, it's really interesting.

And what's really interesting is that the term platform engineering, we live a little bit in our bubble. And maybe people say, well, I know what you mean with platform engineering, but others have no clue. But then if you explain it and say, of course, but we just call it something different. But I think the concept of platform engineering,

of reducing the complexity of the underlying infrastructure to help developers to onboard new applications, push things out into production, get the data that they need in a more self-service way, is basically what everybody agrees on, that this is what organizations are currently building. And different people use different tools for that, they run on different technology stacks. Most of the people that I talk to have chosen Kubernetes as their underlying platform. And then

on top of Kubernetes, they then provide use cases to make it easier for developers to get things out and then they choose the right tools to get the job done. All right, well well let me ask you two questions and I'm going to do an Andy Grabner here and ask you two questions at once and then you're going to have to remember them both to answer. So question number one would be, what are some of the other terms that people are using for platform engineering or conceptual ways of explaining it, but if there are other terms it would be good to know so that if we hear it we know what it is. And obviously with Kubernetes as the platform, what are some of the, I mean, obviously we know we would highly suggest observability, but what are

some of the other tools that are involved in platform engineering? So first with what are the other names for it and what tools do we see in the ecosystem for platform engineering? Cool, I think I can remember both questions. So let me answer the first one uh it's ranging from platform engineering to foundational engineering to devops engineering to just you know building doing cloud engineering cloud architecture so these are some of the things that i hear out there and uh the reason why platform engineering might also be a little bit misleading or like a little strange for some because a lot of software companies,

also like we as Dynatrace, we're building a platform. We're building an observability platform and people build apps on top of our platform. So now if I talk about platform engineering, also internally this could be misleading because, well, what are you talking about? Are you talking about the product?

And we talk typically about an IDP, an internal development platform. An internal development, yeah, I think platform is still the best term here. That is an internal set of tools to allow developers actually to get their stuff out. And I think IDP is actually a good term to be used, internal development platform. The players or let's say the tools, obviously different sets of tools for different capabilities that such a platform needs to have. We saw, at least at KubeCon here,

a lot of vendors are hooked up on Argo. Argo seems to be, at least on the continuous delivery, on the GitHub side, Argo CD, they make the most noise, at least in a lot delivery, on the GitOps side, Argo CD, they make the most noise, at least in a lot of organizations. What are they, like a pipeline tool or like a CD you're talking about?

Argo CD I think is the leading GitOps delivery tool where you point Argo CD to a Git repository, and the Git repository basically holds your desired state. So the Git repository holds all of your Kubernetes configuration, all of your deployments, your

increases, all of it. And then Argo is then using this and making sure that these changes get applied to Kubernetes so that the desired state that you have specified in your Git becomes the actual running state in Kubernetes. And the way it works, Argo and Flux is another tool that comes to mind. All of these tools have basically an operator approach where they basically continuously reconcile and making sure that the desired state in Git is the same as in Kubernetes. And that's the kind of the idea. So as we see Argo a lot,

and what was just an observation, there were several new vendors that just use Argo underneath and then on top, build some additional capabilities like multi-cluster support for Argo management components on top, some dashboarding, some observability, security baked in. And yeah, that's what I saw. So that's Argo for delivery, I would say. That's a big point. Then other tools that people may need to put into their platform is kind of like a user interface

for the developers on top. And that's where Backstage comes in as a product from Spotify. Spotify is open sourced Backstage and it's kind of like a in the end it's really simple it is a service catalog that shows you as a developer what is out there in my organization what services do we have how are they connected it also provides a templating engine where you as an organization can say to my developers, I want to make it easy for them to build a new Java app and get it deployed on a cluster. So you can provide a template and you click that template and then it automatically creates for you the Git repository with all of the boilerplating, with all of the template, with all of the scaffolding so that as a developer, you only need to put in your code.

And then the pipeline comes with the template, all of the observability configuration, all of the security configuration, all this comes with it. So Backstage is one of the tools, but there's so many other vendors that I've seen. Port was another one. There was Deftron I talked to.

There was just like a ton of tools that are kind of trying to provide that user interface into that product, what the internal development platform really is. Because you're building a product, an internal product, that enables developers to get things done in an easier way.

And so, yeah, but backstage, it's just fresh on my mind because Wolfgang Gottesheim and I both work at Downageways, we had a talk at BackstageCon, which was one of the co-located events at KubeCon on Monday, and we talked about how we as Downageways internally adopting Backstage. So that was a good interesting thing. What else? Observability is a big topic, as always and has always been. Are there any specific, you know, obviously there needs to be some sort of a pipeline tool running

and all that. Are there any, you know, we know Jenkins used to be a big favorite of everybody's. Is Jenkins still primary one? Are there anybody else starting to show up? Not as many, I would say. So Argo, what I mentioned earlier, and Flux, these are all kind of like GitOps deployment tools to deploy. But the CI part, still Jenkins quite a bit. And I think they also just announced that

they are now having a SaaS offering, if I'm not mistaken, on Jenkins X. But then there's like,

you know, the Git Labs of the world, the Git Hubs of the world, there's some other tools, Harness

comes to mind as a commercial vendor. Yeah, there's just like a lot of tools that also grow into

both directions, CI, NCD and more and more they are adding additional capabilities like feature flagging management and just rolling out software in a modern way. And yeah, so it's a very it's an exploding market

almost because there's so many new tools coming up all the time. And it's a little overwhelming also if you walk through the expo area. And it's like some of them it's really hard to differentiate. Why would I pick you? What if I have others that have the same message? That's always the problem. It's funny because as you're describing this expo,

it reminds me of, obviously, I do a lot of stuff with music and junk. There's actually one in Germany that happens yearly. I forget the name of it, but there's a lot of these synthesizer conventions and people bring their new gear and their new just experimental things out. And a lot of people go and record bits of it. And you could see a lot of new groundbreaking ideas, new concepts coming in. So thinking in terms of, you know, Kubernetes obviously has been around, oddly enough, we can say for quite a while now, right? We're getting old.

Did you see anything, you know, obviously there's new tooling, but in terms of approaches or new concepts and ideas around the Kubernetes platform, and there I'm using the word platform in a funny way, right? Have you seen anything new where you were like, oh, that's something new and cool that people are starting to do and explore? Or is it just modifications on the existing?

I would say I had the chance, two would say I had the chance to see Kelsey Hightower because he was giving a talk at one of the booths and he did a book signing. And he was asked the same question. He was asked, do you see new workloads being, or new use cases being implemented? And he said, because he was talking with the database booth that were selling storage systems and he was saying, you know, what the important piece is that like

databases, you may not think about the database as being perfectly suited to run on a Kubernetes cluster. That's true. But if you are actually making a step towards Kubernetes and figure out what is the benefit of running a service on Kubernetes because you have the reconciliation loops, you have the capability to declare what you want as a COD, as a custom resource definition,

then it gives some examples of storage systems that instead of trying to just lift and shift the database into the cluster, they actually walk towards Kubernetes and basically said, okay, what do we need to change from the architecture to really fully benefit from the core principles of Kubernetes and the reconciliation loops, the scaling. And that's why I think we may see some additional use case where I think, well, this would never make sense on Kubernetes, on Kubernetes. On the other side, he also said, well, everything has to be on Kubernetes, right?

That's a very clear point as well. The other interesting thing, and I love this analogy where he said, you know, every 10 years, we kind of have this like, kind of like a kind of reset of technology skills that you need. Because if you say, hey, I have 20 years of experience in networking, but you never heard about Kubernetes, then it's about time

to probably relearn. He said, every 10 years, we kind of need to reset and relearn. And then he gave some really interesting advice and he says, hey, if you're an operator and you have your bash script created that you created because people were always asking for the same thing. So you started to create a bash script and then maybe you gave developer a chance to ping you on Slack or maybe create a ticket and then you called the bash script to be a little bit more efficient take this bash script in kubernetes put it into an operator and hook it up with a crd because what that all of a sudden does if you think about maybe creating a new

database instance if you have a bash script for that well you can put this into an operator and you can say i now give my developers the chance to just create a crd that might be called my custom database instance with some parameters, the name, the size, and then it would basically trigger the same bash script in an operator. So I think that was actually a really interesting thing. He said, you know, just try to think what you know already about how to solve certain problems and then actually leverage the core principles of Kubernetes where you can write an operator that is there to listen for the desired state and then do whatever it takes to bring that desired state to life. And so that was really interesting

because we talk a lot about day two operations even where you are automating things that you want to execute in case things go wrong. So you can take your knowledge and put it into an operator. And it's super simple to write an operator even though you don't have any coding skills. So that's really, really, really interesting.

Go on. Because I have another topic. I was just going to talk about the database one, right? Because that got me thinking that databases are hardly ever touched, right? We had some NoSQL many years ago at this point, right? We had concepts like Couchbase and Hadoop and some of those components going on. But as you say, when it comes to database, yes, a lot of the cloud providers offer a cloud version of a database, but it's still this monolithic piece. But the idea of a cloud-native database where it's scalable, deployed, maybe even, you know, because that's the hardest nut to crack, it seems, right?

Because it's so centralized. You can't have locks going on. You can't have the duplicate data or you can't have the duplicate data, or you don't really know. It'll be really interesting to see if something more comes with that instead of just lifting and shifting if there is this concept of a cloud-native type of database that can really operate in the same way as the microservices that get sent up. Yeah, I'd just love to see what comes of that.

It's where we see the least innovation, right? But I think it's purposeful because it's hard to innovate that, right? Because it's like the anchor, it's the core of everything. So how do you mess with that? So it will be interesting to see maybe if something, some mold gets broken on database with that push. I'm not sure if I agree with you when you say

there's no innovation because I don't know, to be honest. I'm not actively looking into that field. That's why I cannot confirm or deny what you just said. It looks like it from the outside maybe. But anyway. If there is is it's underground you know what I mean and my stupid musical thing

people might talk about how the Beatles broke new ground with all this stuff there was crazy stuff going on in the underground it just wasn't on commercial radio so yes the Beatles brought it to the popularity or to the people but there is probably hopefully stuff it would be just nice

cool to see when it bubbles up. So laymen like you and I, who don't see the database all the time, go, wow. Yeah. You said you had another topic that you wanted to bring in. Yeah, because on the keynotes that they did at KubeCon, one big topic was environmental impact, carbon impact, or actually efficiency. In the end, it was really about efficiency. And there were some really interesting statements being made that obviously several thousand people here at KubeCon, many more virtual,

and kind of saying, hey, we are the ones that are building these new platforms, that are building this new software that runs on Kubernetes, and thinking about everything that happens in AI and the amount of compute it needs, we need to make smart decisions with every code line that we write, with every configuration, with every operator, because we will have a major impact on the energy consumption worldwide that our data centers need to run in order to support our workloads. So there was a big emphasis on efficiency, on green engineering,

and that was really good. They mentioned Kepler as one of the projects which I've gotten on a while ago because Henrik pointed me to it. And then if you remember the latest podcast we had on green engineering, it was also Kepler.

It was a very popular product there. And Cube Green and others. So I think the message was very loud and clear that we need to be very careful with our resources. We need to optimize and we need to think about what we do. Because if you build something that then completely takes off in the cloud-native world and everybody's using it, but it was inefficient from the start,

then it will have a big impact. So you need to think about the efficiency aspect. Yeah, and that's a really good point. And I don't bring this up to bash Cloud Foundry in any way because I'm not taking that point. But with platform engineering, I always make the call back to Cloud Foundry of like, they had, you know, there was that concept there in some respects, and now we're all moving back to it.

But if I recall, you needed like something like between like 10 or 20 servers, just, I forget the exact number, but just to run that core of it, right? Just before you even have an application set up. So if these things that we're creating and coming up with require a ton of compute just to even be up and running to serve it, we could look at blockchain and the big problem with Bitcoin mining and how much compute is needed. People have all these server farms, and it's the disaster of that.

We don't want to turn our efficiencies or the cool things we're doing in the compute world into these disasters. So I think it's really cool that the community is getting out ahead of that. And I'd love to see as I think I joked the last podcast we'll know it's taking on or it's catching on when we see that in the North American market. Obviously, who were we talking with last time?

They were talking about the data centers up in Northern Europe with a lot of solar power and swapping around. So, yeah, hopefully that continues, that trend. It's a really good trend to see happening. Because we do, we spend so much time on compute.

And even if you think about it, not to go off, but even if you think about not even just running this stuff, but charging the phones that we use to interact with it. It's just non-stop. We have to figure out a way to

clean that up. Yeah, and that was a big topic. And up. Yeah, and so that was a big topic. And then, obviously, the topic that makes a lot of sense for us is the whole observability topic. Open Telemetry is, I think they just released, like, officially went GA with version, like, it's now GA version 1,

and supporting metrics, logs, and and traces and a lot of vendors. Like I'm always surprised to still see new vendors coming into that space, even though I thought that space is now overcrowded anyway, but there's always room for new ideas. And so that's going to be interesting to see how that evolves. This time was a little bit less as compared to Amsterdam. At least my perception, maybe I'm seeing this wrong, but EVPF was a much bigger topic in Europe I thought. That it was here, at least I didn't see it that much. Yes, I mean, we saw it with some vendors on the observability and security space,

but it was not as prominent, I thought, as half a year ago. Is that because it's not being, do you feel like, is it just becoming part of tool sets? Or is it not, like, you know, if it's becoming more common, it might not be talked about as much. So are we seeing eBPF being baked into a lot more tools just naturally so it's not being considered or talked about as much or do you think it's just not being used as much from, again, this is your perspective?

My perspective is probably more the first thing you said. It's probably just there but maybe people don't make such a marketing fuss about it anymore because people don't know what eBPF is maybe. But that's, again, my perception. And as we said in the past, we all live in a little bubble.

So there was the bubble that I interacted with. But yeah, what I am really excited about, and I'm here in Chicago, it's the end of KubeCon, and I will be jumping on a plane to go to Texas and doing another tour, visiting our community, our Dynatrace community and giving them some insights on platform engineering

in the context of observability. So looking forward to that, meeting more people in the South. And yeah. Where in Texas are you going to be? Houston and Dallas. Okay. Exactly are you going to be? Houston and Dallas. Okay.

Exactly. I've only been in the Dallas-Fort Worth area and Austin but never Houston. Yeah. Yeah, I'm looking very much

forward to that trip because on the personal side my brother is going to join me because he has never been to the States

and we've tried 10 years ago when he had his birthday but then a snowstorm in Frankfurt kind of cancelled his trip and so now we're trying again. There's no snow at least no predictions tomorrow because it's flying over tomorrow.

Yeah, so 10 years later we're trying it again and it will be nice showing him a little bit of the US and we do Texas and then at the end we do Boston. Awesome. Yeah. Yeah, but other than that, I think, Brian, I would have loved to see you in Denver.

I hope we'll get to see each other the next time when I'm there. Yeah, I have to still find a way to get to Austria. I mean, I know I could take a plane. That I know, but like an excuse. But yeah, it always gets tough with the one kid. But yeah, so I'm glad you had a fantastic time at KubeCon. Glad to hear you got to hook up with Kelsey and pick his brain a bit. He's always got plenty of great ideas.

Any final thoughts on that, on the future of Kubernetes? Is there any, were there any, I know this is a long shot, any rumblings of something, the next big thing, or any hints, or is it just all focused? Like, obviously, people wouldn't be touting it, but are there any undercurrents of like... No, I think what it is, is Kubernetes is the standard, right? I mean, there's no

way around Kubernetes, at least not for the foreseeable future, and that's why we have a very strong ecosystem and now it's really, I think, time for organizations to leverage that platform, even though many are still in the starting phase and it's very complex and

a lot of people that I talked to, they were just like entering the field of Kubernetes and they are asking all the basic questions, which is fair to ask, but it also still means that we still need to educate. We need to make the tools better. And yeah, I mean, this may be from a Dynatrace perspective. I like that because we announced

our new Kubernetes observability experience where with the new app that we announced and that will build GA soon. It's about an opinionated view on observability on Kubernetes where we bubble up the right data to reduce that noise and complexity that the Kubernetes cluster generates and really give you the data that you need to get your job done. So, no, but I don't know. I think Kubernetes will stay there for a while. Don't see it go away. and really give you the data that you need to get your job done. But I don't know.

I think Kubernetes will stay there for a while. I don't see it go away. It seems like it's entered the refinement phase, right? Because there's been so many different angles people have taken towards it and so many bespoke versions of it. It really feels like it's reached the point of its maturity cycle where there has to be some refinement. There has to be some, you know, not hard standardizations, but some areas where, okay, if you need this variety,

here's the recommended set of variations instead of just a million in one of them. Because that then becomes really hard for the community to stay on top of. You move from one job to another, and it's another billion different variations. So I think it definitely seems to be in that pseudo-standardization. Obviously, one of the benefits of Kubernetes is the openness of it. You can cater to what you need. But when you have too many choices, there's almost some paralysis or how do you make the right choice? So hopefully we start seeing some of that

maturity coming into it and just the growth into a much easier platform to get into. Awesome. All right, Andy, it was a pleasure having you as a guest. I forgot to play the lead today. See, I came up with some questions. They weren't all dumb questions like I do in the normal

episodes and well safe travels to you tell your brother that I said hi tell your brother that our entire audience says hi and that we wish him well and hope he enjoys his time in the United States and yeah

hope you get back home soon, because we don't want you here now. By the time when this airs, I should be back home. That's kind of my plan. Awesome. All right. Well, thanks for joining us. Maybe we can have you on again in the future, if you

happen, as a guest that is. Do some more stuff and again thanks to all of our listeners any last thoughts any shout outs no okay no just we won't belabor it oh actually one shout out thank you to the local Dynatrace team here in Chicago who invited

me and some other folks who were here for CubeCon to the Diwali party that they had last night that was really nice thank Thank you so much. It was fun. Awesome. All right. Bye, everybody.

Thanks. Bye-bye.