Kubernetes Automated – Cloud Native Now Podcast EP14

Hello, and welcome back to the latest edition of the Cloud Native Now podcast. After a brief hiatus, we're back with Paul Nati, and we're talking about what went on at the recent Microsoft Build Conference, especially as it relates to Kubernetes. Paul, how you doing?

Good, Mike, how you doing? I'm good, I'm good. You know, so all the talk at Microsoft Bill was predictably about ai, and it was a lot of chatter about DevOps and app dev, but talked somewhere in the corner, was this whole thing about automatic, it appears to be a new approach to how we're gonna manage Kubernetes clusters.

And there was a whole bunch of other things that they were talking about in that regard as well. Can you stitch this all together? Because we see a lot of AI workloads running on Kubernetes, but everybody's been complaining about how hard it is to manage Kubernetes for as long as I can remember.

Yeah, Absolutely, Mike. You know, I mean, Kubernetes has been a challenge, you know, and as, as anybody that's listened to any of my recordings or, or been in dialogues with me, I talk about the context in, in my practice of past, present, and future. You know, with the heritage applications moving to containerization and microservices, as well as looking at future kind of technology stacks that's gonna help you with your cloud native development.

That could be like wasm or WebAssembly or, or, um, you know, or, or serverless as a, as an approach. But I will tell you, I was at Microsoft Build excellent, high energy, great event. Um, there was a lot going on there that was focused, as you said, focused on the DevOps teams, focused on developers.

Of course, as you would imagine it was a developer focused conference. But I think the big thing that, that was really the takeaways from, from my perspective, was AI everywhere. Sure.

That's a, that was kind of a given. We expected that. Um, but really trying to understand co-pilot integration across on-prem, across Azure, uh, Azure development tools was really a big focus area.

We saw Microsoft co-pilot integration across, uh, the portfolio. It really does become, and I don't want to, I don't wanna over expand on this, but it does become a game changer for developers, right? It, it because it, it allows developers to really, uh, infuse AI into their workflows and, and, and productivity tools.

So they can actually focus on innovation rapidly, right? They can drive results and drive towards those, you know, con contextual suggestions, code, uh, recommendations, and also best practices. You could put governance and regulations in place.

So I was really excited to see that. So, yeah, what's happening there? Um, there's a lot more to talk about there, Mike, but what are you, what are your thoughts on, on how that kind of came together?

Well, I think what we're seeing is a, is a two step towards the democratization of the management of Kubernetes itself. I think the first step are these platforms like Automatic, and there's other ones that are out there, but, um, they make it accessible to your average IT administrator. You don't have to be a DevOps engineer necessarily to manage these instances of Kubernetes.

And you may not even know what the underlying API ever looks like. I think that's step one. Step two is putting these copilots in front of, uh, everything, including Kubernetes.

And I think one of the things we saw was there was a copilot for Azure that now speaks to the Kubernetes clusters as well. And that will enable folks to diagnose and troubleshoot Kubernetes clusters pretty easily. I think, you know, I think you might still wanna know what's happening underneath there at some point, but, um, you don't have to be a rocket scientist to maintain this environment.

We can have a scenario where the day-to-day management is handled by IT admins, and then, you know, if there's trouble dial 9 1 1. DevOps. Yeah, I, I think there's a little bit of that.

Um, but, you know, I will say, I will bet anyone, the audience that's watching this, anyone to a shiny quarter, if they can find me somebody that still calls themselves an IT admin or especially at one of these events. Everything, uh, every event that I've been to over the last, say, call it 18 months or even more, has been, you know, the, the, the, the role of a dedicated resource. So like a server admin, virtualization admin, network admin, they've all kind of morphed to the platform engineer or the DevOps team.

So, so I do, although I do agree with what you're saying, the democratization of AI tools really was a big focus for, uh, at the event. Uh, and the copilot integration really does help those enterprise developers and makes it easier, right? To, to harness that power of those AI projects, uh, across Azure, across, uh, across the, you know, on-prem heritage applications as well.

So there is that piece of it, but, but the thing that's was most interesting to me was, um, when you look at the democratization of these tools, there was an high emphasis on efficiency and security. com that really kind of highlights a lot of the findings and why Microsoft build, you know, 2024 really did change my mind, right? What I changed my mind about what's happening.

I, I kind of did go into it thinking, all right, well, we are gonna talk about ai. Everything I did, I sat through about five or six different demos around copilot and the integration on the different areas, whether it was, you know, uh, Microsoft 365 or, or the, the, the pc. So there's a lot happening there.

So there's a lot of harmonization across the development. So whether you're on, um, you know, whether you're on-prem or you're on, uh, building your, your sandbox environment and then deploying it into the cloud and Azure, uh, there was a lot of kind of, uh, synergies and made it simpler. And it was very clear to me, Mike, that Microsoft build the message was, here's the tool stack, but it was also addressing the skill gaps issue that were out there.

You know, it's a good segue to what was being discussed at the Nutanix conference, which, um, occurred in Barcelona a couple of weeks back. Um, and it was interesting 'cause I was talking to the CEO of Nutanix about this very issue involving quote unquote IT admins and platform engineers. And he was the opinion there's a lot of upskilling still required just given these folks a bunch of tools does not magically turn them into platform engineers.

Um, and it seems like there's a long road to go there. Many of them though, he was likening to the ship to becoming, you know, VMware specialists where they all kind of eventually migrated up from, um, managing lands or whatever it was, and all got a raise 'cause they were able to call themselves VMware specialists. And I think the same thing will happen with platform engineering.

It's not clear to me that these people are actually quote unquote engineers. I mean, they got a certificate somewhere, but, um, you know, does that really make you a, a, a certified engineer? I don't know.

Yeah, you know, it, it's interesting. So, uh, look, I mean, Nutanix has done some really interesting moves, uh, strategic moves, I would say in the, in the, in the multi-cloud, distributed cloud, multi-platform, uh, environment. Project Beacon really kind of speaks to this, right?

Uh, my colleague Guy Curry was at, uh, was at Nutanix next, and we were very, we were covering the event very closely together. This, again, a a number of research notes that we published on, uh, Nutanix, uh, uh, platform, um, the new Nutanix platform, our NKP. One of the things that was interesting about the platform is I've been, uh, working with, um, D two IQ for quite some time, and the, the acquisition of D two iq, really, uh, Toby and the team there really have brought a lot to the Nutanix, uh, portfolio.

So the acquisition that that brought in D two IQ into, uh, into Nutanix, really helped with that Project Beacon kind of approach of that distributed cloud distributed multi-platform, uh, you know, database data services and storage databases, you know, regardless where they're hosted either on-prem or in the cloud. So there's a lot that's happening around that acquisition. I'm really excited to see where it's going.

Um, there's definitely some harmonization that needs to come together between, uh, the, the D two IQ tech stack, uh, and I keep calling it D two iq, but the NKP, which is what was announced at, at, uh, at the end of next conference, uh, as well as NC two, right? There's NC two, the NC two piece, that, that also has to be, uh, taken into consideration. So whether you bring your own platform or, you know, you has the ability to kind of, uh, adopt or, or migrate later, that's a, that's another interesting kind of approach.

So I like where they're going of meeting the client where they are. I like the fact that they have the tool stack to support meeting the client where they are, but they also have the ability to grow for that, call it the outliers that might be more advanced and more looking at automation and looking at Kubernetes across distributed cloud environments. I mean, we're seeing in our research, 94% of organizations are using two or more cloud, uh, offering.

65% are using four more clouds, right? And if you are using these multiple clouds, distributed clouds, you need a tech stack that's going to harmonize those, those different cloud environments. Yeah.

D two iq day two IQ was, um, I guess what they were trying to drive with that whole thing problem was there just wasn't enough day two for them to kind of sustain the company. And, but now it feels like, ironically that we are seeing enough Kubernetes clusters out there to justify a day two mindset, and it seemed like a good fit with Nutanix, which is trying to figure out, um, how to turn that conversation into something that fits their base. And when I talked to them, I mean, let's be honest, they couldn't, they're just not really a DevOps company.

They're much more of an it sm company with a bunch of servers and packaged around a hyperconverged platform, and that's their base. And they don't mind saying that that's pretty much what they're after. I'm, it was just interesting to see whether or not, um, you know, those folks are gonna take over that role or function, or if the DevOps team is gonna wanna hug those Kubernetes clusters longer.

There's an argument to be made the other way that says a lot of these DevOps folks would just as soon get rid of the infrastructure and focus more their time and effort on application development and delivery and Kubernetes just kind of gets in the way. Yeah, well, I, I'm like, I will say my, uh, my entrepreneur side of me actually says D two IQ really did, uh, Excel because they, they, they were acquired, right? They are now part of a larger company.

So in bringing that tech stack into Nutanix really helps, uh, excel at Nutanix to that, to that up stack level. Now, I will say Nutanix had a tagline for a long time that was the invisible infrastructure, right? And they wanted the infrastructure to be invisible because they didn't want it, it just needed to work.

And, you know, frankly, when you're a DevOps team or platform engineering team, it should just work, right? So if as long as the, uh, appropriate provisioning of the resources are assigned to the DevOps teams or developers or the platform engineering teams that everything's good. The problem is when you leave that control to the DevOps team over provisioning occurs, right?

Because the DevOps team looks like, Hey, we just need to make sure that this is running, and if we need more resources, fine. 'cause they're not paying for it, right? Um, but, but the other side of it is also true that if you can, um, optimize it and make it more efficient in the backend, then as long as the DevOps team platform engineering teams are happy, that's that's great.

And I think that's where Nutanix is going with this acquisition, Right? I'm not entirely sure I agree with your definition of winning unless you're Charlie Sheen, because as far as I saw, they, they sold the assets rather than the company. So, uh, you know, it's, it didn't quite pan out the way they'd hoped, but they have a lot of good customers, and we'll see what phase two of that whole act starts to show up.

Interestingly enough, Rafa Systems, which is a competitor, um, is now talking about how to automate the management of AI workloads running on Kubernetes is an extension of their COP platform, and they too are applying generative AI to the management of Kubernetes. So, um, what's your sense of where do they fit in the spectrum and how many AI workloads are out there and what makes 'em different? Yeah, Sean and I have had many, many conversations about where FFA is going and where they're taking the company.

Um, I think it's the smart move, obviously, that they're moving into, um, uh, to kind of a layered approach, right? Going in and applying AI into their Kubernetes workloads allows for that automation allows for that, uh, actionable insights based on what's, what's occurring for the workloads that are, that are being built on top of it. Um, they had to, right?

Because if they didn't do that, uh, all the other players around them in the orchestration space, um, would've basically come in and done that anyways, right? So they had to add in, uh, the actionable insights they had to add in the automation as well as add in the, um, the AI functionality. The other side of it is, is, you know, they, they definitely have a vision of moving, kind of, uh, making it more transparent, making it more usable for the, the operation side of the business.

And I think that not just from a day two perspective, but from a, a day zero and day one from a build and release operational side, that, that also kind of plays into it. So I, I'm, I'm impressed with Ray Fey. I'm, I'm looking forward to seeing where they're going, um, because I think that they're, they're on the cusp of doing something really, really cool, and I think that they need to kind of, uh, they, they show that they're gonna take that to the next step.

What is your sense of, uh, where will these AI workloads kind of fit into the existing management workflow? It used to be kind of the, you know, there's an ML ops team and they built the model and they manage the model, and, you know, they kind of exposed an API and hope for the best. It seems like these AI models are kind of now being embedded more into the applications themselves, and there's inference engines and they're gonna be deployed everywhere from the edge to the cloud.

Who's gonna be in charge of kind of managing the deployment of these AI models? Yeah, that's a great question. You know, I, I, I think the way I would view it is, um, a year or two, maybe a couple years ago, I, I built this application maturity model, right?

And this application maturity model was about five phases, and it talked about how you get to your heritage applications to a fully elastic cloud native environment and, and then what that means within organizations and how organizations are using those applications to kind of grow. Now, what I found is, you know, when I first started doing the research, and I started doing the maturity and modeling and, and kind of understanding it is, you know, I thought that there would be far more, uh, organizations that would be like to the right, you know, like more to the mature side of it. And I, what I found was just the opposite as it was, there was far more organizations to the left.

And so when I look at those phase one, phase two, phase three, um, and up to phase five, that that starts with heritage applications to maybe virtualized applications to, you know, maybe a fat container, maybe a, uh, microservices and then a fully elastic environment. And, and applications can sit anywhere within any of those stages, and they could be multiple applications within organizations. So the way I see AI workloads kind of tying into each one of these applications or these phases and this maturity is they will grow with the, as, as an enabler to that application within its own pillar or its own silo.

So if you have a heritage application and you have say, we'll just say you have a heritage application that uses a, a, a, A data lake, okay, that data lake, that data needs to be, you know, that is now a new, it's a new approach to looking at the, uh, the data sets, right? But that new approach is not anything new. That's what people were doing previously.

And now, uh, those, you know, those, those data sets are going to be maturing as people, uh, mature those applications in their organization. So I think it's an evolution. I don't think it's a one and done, and I don't think it's one work, one group is going to own it and another group is not.

I think it's going to be an enabler for the, for the success of the business. Yeah. I scratch in my head about the quote unquote groups, right?

I mean, it takes a frigging village now to do anything. I gotta get a ML ops team that work with a DevOps slash IT team that's also gotta bring in a bunch of data engineers, and then I gotta throw in some security people. It's a wonder anything gets done.

Um, do we need to kinda restructure the way we think about how IT teams are organized? I mean, restructuring the way IT teams are organized, I think we did that. Uh, we have been doing that for years, right?

I mean, remember the days where it owned everything? You couldn't do anything within an organization without it's approval. Then we over pivoted and like said it does nothing.

And now it's only going to be in the lines of businesses, which created all sorts of craziness, like shadow it and just organizations just standing up their own things. And, and it really wasn't a good, good model. But then we started re internalizing it, right?

And saying, okay, it is no longer we're gonna stand you up and you're gonna just run equipment. You're going to be business management leaders, right? You're going to look at what tools are needed, how to put the governance and controls and guidance in place in order for it to work.

So when I look at it, it over, I, I like to, I like to describe the persona kind of in two roles, right? There's IT and this ot, right? The operational teams are the ones that are across multiple facets of the organization.

So you can have lines of business, you can have, uh, the, the, you know, the people that, the, the DevOps teams, the platform engineering teams, but then you have it as a whole that may have the budget, that may have the control and the governance, and they set their, they set the tone for the rest of the business. Is there a pivot? It depends on the maturity of the organization.

I mean, I find, I find a lot of organizations that have IT lines of business, it DevOps lines of business, IT DevOps, SREs lines of business and IT, and platform eng engineering and lines of business. So, and that's not large and small. That's across the, across the, the whole ecosystem.

So Yeah, I think the trouble I have with this whole platform engineering concept is it kind of sounds a lot like the revenge of centralized it. The reason we created DevOps in a lot of these app dev projects was to get out from under the thumb of folks who, um, you know, worship a little bit too hard at the altar of idol and everything takes too long. So how do we get to a middle?

Yeah, I, well, and I think that's a fair point. I think that the one caveat I would add into it is the focus is no longer, um, on the infrastructure, no longer on SLA uptime for infrastructure. The focus is on SLO and why it's important for that business logic to be running, because your infrastructure can be running, but if your applications are down, you're still down, right?

So that the platform engineering teams kind of are focused on the running of the business logic versus running the infrastructure itself. So I think that's where there's a little bit of a pivot, but I don't, I, I mean, I agree with your, your assessment that there's maybe a little bit of control, a little bit. Um, but I think organizations need it, especially when you start introducing the, the rapid change with AI and the rapid change of actionable insights, you need to have governance and control in play.

All right, folks, you heard it here. Well, you know what? Change is coming and it will be different in every organization, but whatever you were doing yesterday, it won't be the same thing tomorrow, for sure.

Hey, Paul, thanks for being on the show. Thanks, Mike. All right.

Thank you all for watching and listening to the latest episode of the Cloud Native Now podcast. You can find this episode and others on our website. We invite you to check them all out.

We're also on Spotify and all the other popular services. Until then, we'll see you next.