KubeCon + CloudNativeCon EU: Wasm, AI and Rancher – Oh, My! – Cloud Native Now Podcast EP8

Hey folks. Welcome back to another episode of the Cloud Native Now podcast. I am Sharon Florentine.

I'm here with Mike Vard. Although I should probably start by saying Bonjour, right? Although my French is non-existent.

I took Spanish in school. But, uh, how's Paris living the dream? Paris is lovely.

We're here at the CubeCon Plus Cloud Native Con Europe conference, and as usual, it's been chock full of stuff, so we can jump through, but there's no shortage of things to talk about, that's for sure. That sounds great. I am, I've been jealous all week, so I can't wait to hear what you guys have been up to.

All right. Was has been a huge focus, which is something we've kind of anticipated over the last year or so, right? Yeah.

Let's jump in. Well, we'll start with Wasm. Um, it's interesting times, right?

It's still really early for this, uh, web assembly stuff, and it's a file format that lets you, um, at least make it easier to build applications that can, you can write once and run anywhere, right? This is the original promise of Java. Finally realized some decade and a half later, 30 Years later.

Yeah. Um, but it's interesting conversation because, um, here at the conference, the Cloud Native Computing Foundation has essentially two projects going on, but that are, one is focused just solely on the runtime environment for, um, wasm, which, you know, they're arguing should be Kubernetes. And so there's an outfit called Onic that built Wasm Cloud and that helps you deploy that.

The other is Ferion, which went a little step, maybe wider, and includes a development environment and deployment and management framework for managing this stuff end to end, to create more of a serverless experience for Wasm that sits on top of Kubernetes and they donate, or at least applied to donate their Spin Cube project to the CNCF. And we'll see how that plays out. But the other thing to keep in mind is there's this other group out there called the Bike Code Alliance that really drives wasm itself and they have both tools and projects for building apps and runtimes as well.

So it'll be interesting to see how all this plays out between these two different consortiums with, uh, different projects going to different places. The upside is, um, the hope, hope is we're gonna have a better developer experience for folks who are riding on Kubernetes platforms. 'cause today it's, uh, shall we say, heart and uh, as such, um, you know, the number of developers that are writing to Kubernetes is probably somewhere south of 10 million.

And there's probably on an estimated 50 million, uh, enterprise developers out there. So that's 40 million people who are going, no. Uh, yeah.

And so we gotta make this a little more attractive to those folks who are used to frameworks that let them write at a higher level abstraction. And they don't really wanna know so much about the underlying infrastructure. I think, um, even the 10 million or less that are writing to the platform or doing so under some form of, shall we say, duress because there's a DevOps team somewhere that said, thou shalt and we're standardizing on this.

'cause it makes their life easier, right? And, but, um, so I think we have to get to more of the coalition of the willing, shall we say. And, and so part of the conversation here is how do we kind of do that?

I think Wasm ISS gonna be a piece of that puzzle, but this, uh, ongoing tension between developers and ops folks around Kubernetes has been with us for a little while now. And as far as I can see it is gonna be with us for a little while longer. Yeah, it seems that way.

And, uh, to your point about the tension between the two competing factions, uh, we had a piece a month or so ago, postulating that Wasm might be, you know, the next evolution of what Kubernetes is supposed to be. And it definitely sounds like we are not there yet. If you listen to the Wasm crowd, they're saying, you know, faint praise for Kubernetes, but they're also pointing out that, you know, laws can run anywhere.

So you don't necessarily have to have Kubernetes. And a lot of them are kind of, you know, wink, wink, nod, nod going, well, maybe we'll be seeing these containers in Kubernetes platform some other place some other time, and we'll do something else. But it's not clear to me what that's something else is just yet.

'cause um, you know, you theoretically you could build a monolithic application, but, you know, microservices have some benefits that, um, make it easier to maintain apps over the long haul and apply rip and replace kind of patches. So, um, we'll see how this all plays out. There's still a lot of developers out there resisting microservices though, because they prefer monoliths because, well, from their perspective, they're just easier to work with.

So, um, yeah, we got a lot of, you know, competing visions for the future. Yeah, indeed. Uh, that brings me to another point.

There was an, uh, end user technical advisory board. Am I, uh, am I correct there that the CNCF announced today? How do you think that ties in with the complexity that folks are experiencing?

Seems like that might help developers cloud native or otherwise at least be able to wrap their hands a little better around how to actually implement this stuff and get over that hump. Maybe ease the duress a little bit. Yeah, I think there's a lot of things that work here.

The board, which covers more than just Kubernetes, but has, you know, so hopefully will be applied across, I don't know, it's CNCF today. It's the, the landscape has hundreds of projects in it. It's, uh, hard to keep track of all of 'em.

Not all of them are all that mature. In fact, the ones that we call mature are still considered immature by others. So, um, well, you know, there, there's clearly work to be done to, and I think part of the issue is the feedback loop has been somewhat broken.

Um, so this board will create a loop for end users, including folks from Apple and Spotify and all these other folks who are leading edge users of the software will, you know, have a channel to kind of commute, communicate more directly with the maintainers of these projects about what their requirements are and what they need, which should help those folks prioritize some of their projects. 'cause a lot of times, you know, let's be honest, it, there's still this kind of like the maintainers are like, well, I think this is cool, and then they kind of take a vote and away we go. But, um, maybe we need to prioritize certain more mundane things that need to get fixed before we go on to the next big great cool thing that we all wanna work on.

And Right. So a a little adult supervision here, that's what I would call it. And that sounds like it is coming at exactly the right time.

Yes. Uh, maybe arguably overdue. That's Yeah.

Or that I was trying to be tactful, so, all right. I've gotta bring the AI question into it. Uh, what's going on with, with AI that you've seen at the conference?

Because that was kind of the, that was gonna be the huge focus. Did that play out what's going on? Yeah, everybody was talking about ai.

You almost think that there was nothing else going on in the world, but, um, Who knew the Yeah, but the issue is the data science teams that build these models are encountering the same issues that developers are encountering, right? They're like going, this Kubernetes thing is hard. And so if you look at the major AI platforms run by the big cloud service providers and other entities, you know, that have hundreds of software engineers, um, they're all running on Kubernetes because, you know, it's a pretty ideal platform for that.

But, you know, as the one of the fellows from Microsoft here point out, we gotta make this easier, it's too hard. And if we don't make it easier, then we won't have enough models that people are willing to build and deploy because, um, there's not enough engineering talent to go around. So, and we're kind of having this DevOps conversation all over again, right?

We're trying to figure out how to make the data scientists play nice with the developers who in turn play nice with the software engineers. So in turn, talk to the IT operations people so that we can actually build and deploy these things in some sort of workflow. And I think I heard the word empathy, which is, you know, DevOps code speak, um, maybe at least a half a dozen times over the last two days.

So we're all trying to reinvent the wheel and some people are calling for a new job title, surprisingly call the AI engineer. Okay. And if you swap out AI for DevOps, it's pretty much the exact same job.

That sounds very, very familiar. So along with the cutting edge AI stuff, there were some Docker announcements and that's kind of throwing it back right to the, the very beginning of containers and the spark that ignited cloud native. Yeah, It was kind of funny to watch, um, Docker and Red Hat get together and agree on something 'cause you know, for so long that they have not always seen eye to eye.

But, um, there is an agreement that, uh, red Hat OpenShift will, uh, support this, uh, test container cloud thing that Docker has, which is a cloud, it's a framework really for, um, testing applications that'll accelerate that process and enables it to shift a little further left. But, um, it was originally developed by Atomic Jar and they got acquired by Docker. And now that's being brought together inside of this service.

And I'm hopeful that maybe this will be exposed to other platforms as well. Ideally, we want, you know, the applications to be higher quality. And the easier it is for developers to test stuff, the more likely that will occur.

And I think we forget sometimes with testing, it's like it's hard to create the test. It takes time. And if I'm behind schedule, you know, the first thing I'm gonna maybe short tripped a little bit is those tasks because I'm gonna be like, well, maybe somebody else will do that later on.

Or, uh, I will tell myself that we'll fix it on the first patch, update whatever issue comes along. And you know, of course by then I'm probably working on something else and I forgot about it anyway. Mm-Hmm.

And I think we're gonna see more testing. We just gotta find a way to make it simpler and easier for folks to do because no developer gets out of bed and says deliberately, I'm gonna go build a crappy application today. Right?

But, you know, things conspire to kind of result in suboptimal experiences and performance and vulnerabilities and all these kind of issues. And I think, you know, we all should just stop pointing fingers at each other and just say, look, the system is broken and let's go figure out how to fix the system. Indeed.

Mm-Hmm. And, uh, you know, that brings AI back into the conversation because AI is again, being put forth as this is how we solve the testing problem. AI can create the tests, it can run the tests, it can figure out how to fix what the tests flag as flaws or vulnerabilities.

But again, we gotta get past that initial hurdle of how do we make it work and how do we integrate it into the workflow And so many things. And it will also rat you out when, for the test you didn't apply and the test that you missed or didn't address. Because I can run a report that kind of identifies where there's issues in the code, so eventually you're gonna have to fix it anyway.

But, um, hopefully there's a virtuous cycle somewhere in this whole thing that's gonna emerge. But I'm not quite clear how long that's gonna take. Me either.

Me either. And then the most unusual perhaps. But, um, thing that we don't talk much about in terms of technology that was at the show that was worth checking out was LOF Labs, which, um, makes this kind of virtual Kubernetes cluster.

Yep. Uh, They, uh, partnered up with Rancher. And so now I can use the Rancher platform to manage all these virtual Kubernetes clusters.

And the idea has been, you know, to deal with the potential of Kubernetes sprawl, which is all these physical clusters. Um, I can now take a cluster, a big one and share it across multiple virtual Kubernetes clusters. And a lot of folks have been using that in pre-production environments for testing.

And, um, but some folks are now starting to deploy that in more in, uh, production environments. 'cause they're like, I cannot manage all these physical Kubernetes clusters out there, and I need something that I can rip and replace when needed. And it's much more disposable.

And, you know, because a lot of projects are short run maybe, or some things are, don't require as much, uh, processing power. In fact, you know, the original Borg that was at the heart of Kubernetes, they envision, you know, these monster clusters that would be shared, um, versus, you know, the, all these distributed clusters that we're currently seeing. I think there'll be distributed clusters at the edge.

But in the data center itself and in the cloud, I think there's a lot of opportunity to save some costs and kind of not over-provision, Kubernetes clusters. 'cause that's something we all like to do. And of course that has climate implications, but Right.

Every developer I met, you know, is more terrified about the application being down and getting a call at 3:00 AM so they will over-provision regardless of cost and, uh, sustainability concerns because well, who wants to be woken up at three o'clock in the morning? That's, yeah. Do you, do you see more focus on the sustainability or it's still just not there yet?

Well, folks, were talking about it here, and this is Europe where they pay a lot more attention to that issue than you do in the US across the board. So it definitely getting, um, uh, airtime, but I think ultimately it's gonna be addressed as an extension of this finops effort. So now I'm gonna have a program to monitor the cost.

The costs are directly tied to consumption, which is tied to energy, which is tied to how much carbon we're kicking off. So I think that eventually all those things are gonna move into one motion. Uh, it is just a level of maturity and providing the, the, the metrics that we need to track that alongside all the other metrics that we have.

I'm kind of a big fan of, you know, let's make all that stuff part of the DevOps workflow so we can see that as part of what we're doing, versus having this kind of orthogonal thing over here where there's yet another platform that I already, you know, I'm suffering whiplash from, from all the platforms I'm supposed to be looking at anyway. Yeah. And it seems a lot of folks are thinking the same.

Well, maybe, uh, not a lot. Uh, off the top of my head, I have heard about at least one or two observability platforms that are integrating, uh, carbon footprint tracking into their existing platforms. So maybe soon, someday, maybe.

Yeah. I don't think this in itself is gonna solve global warming issues, but it's a big giant step in the right direction. And, um, let's be honest, right, with AI and all this other stuff, we're kicking off more compute capacity than ever assuming more power than ever.

So, um, a lot of this is like, you know, excuse me. Um, I would like to run about 50 more workloads that I'm currently running without getting called out by the climate police. Yeah.

Yeah. It is always a push and pull. Not to be overly cynical about it, but, you know, some of this is just kind of, you know, shall we say, uh, I, I prefer not to be climate chained.

Yeah. Yeah. That's true.

All right, well thank you for the cube con report for those of us who could not attend. And, uh, yeah. What, what do you expect for the, uh, the rest of the show?

Well, the show's wrapping up here, but I guess the next event is in Salt Lake City and then the one back here in Europe will be in London next year. So we'll be looking forward to all those events and, um, hopefully folks, you know, if you go come on by the Textron group booth and say hello. Yeah.

And, uh, come back and chat with us again next week, folks, we'll be here. And, uh, thanks so much VO.