Techstrong Gang KubeCon Paris – March 21, 2024

Hey everyone. We are live in Paris, CubeCon for a very, very special edition of all Things Cloud Native and more on Textron Gang. All right, we're back here.

We're live. If you can't hear the roar of go, what's going on on this show floor? It's because our mics are really good.

But it is a, a raucous, you like that word? It's a raucous crowd here at Techstrong. Uh, at where we're at.

We are tech strong at CubeCon in Paris. And um, it's been a interesting first day, Mike. I know you were, uh, yeah, I had a whole bunch of the, uh, conferences and keynotes.

I Went to the keynotes, the press conferences, and, you know, there were 12,000 of my closest new friends at the, uh, keynote this morning. So that was pretty cool. I think it's close to, well, maybe all, there was 13,000 people here, so Yeah.

Probably 12,000. Mitch and I have been holding down the fort here doing, uh, interviews and, and we've seen a few trends already. I think starting to emerge.

Some of them are probably to be expected, some maybe not. The first thing I wanted to talk about is the, the continuing maturation of the cloud native platform or the cloud native ecosystem, much as we talk about like with DevOps and DevOps next. Yeah.

Right. Where we're seeing less point solutions that you need to cobble together to more cohesive platforms. And it's almost platforms within platforms, kinda like nested Russian dolls.

Right. And, um, you know, I, a lot of the interviews that we did today, Mitchell, that was the, the theme. And it goes with, you know, what, what is tomorrow's, today's product is tomorrow's feature.

Mm-Hmm. And we're seeing that a lot. Yeah.

With, I'm hearing it from the vendors and people we're talking to. You know, it's also another parallel with DevOps is it isn't cloud native, literally microservices and containers and just that, right. Anyway, as narrowly as you wanted to define it, it's the ecosystem of developing cloud applications ranging from OpenShift is cloud native to Yeah.

Kubernetes and these things. But it's also the platforms that support all of that. And I think that's what everyone's realized is this is so complex.

You can't do point solutions all over, up and down the stack horizontally, vertically. It's crazy. It, there's so many variables.

I think we were talking about this earlier, but how much of this show now is it ops people and the back end of the DevOps people versus When we first started coming into this, it was mostly developers. All developers, all developers. That was just a couple years ago.

It wasn't primarily developers, But you know, many of the interviews I did today, it was about taking the load off developers, making it easy for developers, whether it's through using AI or ML or just better design programs for them, was about making life easier for developers. There's a tension here though, 'cause there's an irony in the whole thing. Originally Kubernetes driven by the, so-called full stack developers, of which there are not many.

And then they convinced the IT ops people somewhat unwillingly to embrace Kubernetes. 'cause it has standard APIs, and they finally got it. But now the rest of the developer community is dragging its heels because the abstraction layer is not a lot of fun yet.

It's not. It's, it's still hard. Kubernetes is not easy.

It's hard For the developers. That's the line. And so I think that the next wave of this thing is we gotta get to another abstraction above Kubernetes just that the developer sees.

So they don't see all this underlying stuff. There's a, a growing, uh, wasm community that's kind of banging that drum saying that Wasms gonna be the new front end for Kubernetes, and that wasm will replace containers with something that's more of a lighter weight component model that runs faster. It's all integrated.

There's a couple of announcements around that today and a couple of tomorrow. But, um, that's gonna take a while. Wasm is early.

I mean, wasm doesn't really have a vendor ecosystem that's meaningful yet. It has a couple of players that's Two players And, and it might be two, three years before Wasm is fully, you know, enterprise ready. It'll be longer than that.

You know, speaking, I'm sorry. Again, let's Just say the irony is Kubernetes is, is everywhere. It's not just developers and not, not just ops.

It's in all these products that are being run in the cloud, right? Whether you're running OpenShift, but you're running it on Kubernetes or you're running an object store database that's interfacing through, it's all managed through Kubernetes. So in a way we're, I dunno if we're making a problem harder or easier, maybe we're building up the expertise to operate these things, but it's sort of proliferating the same technology.

I will tell you that there are certain vendors that are, they have a credibility problem, right? 'cause they've been running around saying that we abstract Kubernetes and we'll create the development environment for you. And yet you talk to every developer out there and they're like, that stuff's too, too complicated.

So all that talk about developer productivity, cloud native and those platforms, something's not connecting. There's a miss, there's a piece missing. So, you know, you talk about containers and Kubernetes.

I spoke with Scott Johnston, the CEO of Docker about this. And I will tell you, they, you know, they, I mean Docker pivoted in 2019. I think we all know this to really focus on developer and they're, they're trying to make it easy for the developer not to deal with Kubernetes, right.

By, by using some of these new Docker tools. Um, I don't know if abstracting Kubernetes out is the answer. I would tell you one of the first vendors to announce support for was and was Docker.

Really? Yeah. Oh really?

Okay. Full circle number One. There you go.

But I, I think what it is though, you know, it reminds me when cloud first came out, right? Our friend, rich Mogul had a term called, uh, uh, uh, cloud Washing Security, which was taking security and just moving it up to the cloud. But it wasn't really designed from the ground up, the can down the ground.

Right? Right. I think what we've seen is cloud native wash developer tools where we, we, we just kind of slammed Kubernetes on and, and all that that entails and said, okay, developer, you gotta develop in this environment.

And now I think companies like Docker and some of the other companies we've interviewed here today are saying, no, that's not the right approach. We don't need to shove Kubernetes down their throats. And they, whether it's an abstraction layer, ai, war automation, what have you, serverless, That's another Option.

Developers wanna write code. Everything you do that takes them away from writing code because they're doing something else or waiting on something else ruins their productivity. So I think that's the focus.

I mean, when I look at Docker build and I listen to them, to me that sounds like a DevOps platform. They're just not calling it that. It sits up in the cloud.

I did, I did speak to him about that. Um, had a very, by another name, enthusi. Yes.

You'll, you'll need to watch my interview to find out. But yeah, you know, they, they are definitely looking to shake up the CICD market to make it easier for developers. Yeah.

You know, another big theme I, not unexpectedly, but it's platform engineering constantly as well as AI kind of the two bookends of every conversation has at least one or both of those in those at least ai It's interesting too that, you know, platform engineering, there's different ways, segments of what we think of that is. One of them is the developer portal, internal developer portal. There's also a school of thought now around, well, that, that's kind of a bookshelf or a shopping shelf for all the prebuilt patterns, things, configuration tools, et cetera.

But there's still, it doesn't eliminate a lot of the work that it takes. It just kind of puts a bit of a garden around of it. And maybe we need more cloud development environments that are kind of built for you to be productive.

Maybe. I was talking to a fellow who was one of these agile framework people and, you know, hardcore, and he was not having the platform engineering conversation. 'cause he was like, the whole point in his mind of agile was, I get to choose my tools, I wanna be like in charge of that.

And now you're telling me that the man is back from central IT with platform engineering. I'm here to help and telling me how I'm gonna like do things. And he is like, thanks, but no thanks.

Well, Standing behind him was the, uh, the waterfall person, right? Like, well, wait a minute. Yeah, maybe.

So yeah, that, you know, I always try not to discuss Politics and religion. Anyway. Hey, let's take a break.

We'll be back with Textron gang in a minute. com is the number one online destination for DevOps education and community building. com covers all aspects of DevOps, including DevOps, best practices and tools, DevOps culture, DevSecOps, business impact, continuous testing, continuous delivery and more.

com has the largest collection of original DevOps content featuring breaking news, blog posts, podcasts, and more. com to learn more. com where the world meets DevOps.

All right folks, we're back at CubeCon and now we're gonna talk about, well, surprisingly, AI can't go, you can't even leave your house today without somebody accosting you about their AI thoughts. But, uh, here at the show, we're talking about how Kubernetes has emerged as almost like the deep ACTO platform for ai. But I gotta say they were going so much out their way to like say that and to, you know, invite the AI community into the cloud native community that I always get a little suspicious.

And you talk to the data scientists and they're like, you know, they're, they have the same complaints that developers have. They're like, this stuff is hard. I don't want to see this stuff.

And you know, they have their own little ML ops frameworks for building the models and they're like, Kubernetes is hard, but I don't really want an IT ops guy in the middle of that conversation. So, but at the same time, they're not gonna be able to do anything with the without it. So I give, uh, Priyanka Sharma some, uh, props.

She was saying during the press conference, we gotta get that DevOps mindset into this conversation because people have to have empathy what other folks are doing. And, and she was like, we're gonna have to create some new framework. She didn't have a name for it or she had one, but it was kinda like dev ml ops SecOps thing and you just didn't roll off the tops.

See that I can get behind because DevOps isn't about a tool or technology, right? And you're talking about how do you solve work problems, work, work streams, flow of work, and the different characteristics of it. What it like to manage a model or build an LM or to support the data, the training, all that kind of stuff.

Yeah. Those are unique things. They don't literally fall into A-C-I-C-D pipeline flow.

So how do you meld those in? That's not Kubernetes. That's a, that's how work happens.

And how do we build platforms that help multiple things work on their own path and also coordinate where they need to, but people working together. More importantly, I think it goes back to what we said before. If AI is going to make people's lives and jobs easier, if it's gonna free up the developer to write more code, it's gonna free up the ops guy to, to deploy faster and better.

All of the naysayers in the world don't amount to it hill of beans as Humphrey Bogart would say, right? Because it's doing what we needed to do. Will it do that?

Is I think the $64 million question. Uh, I'm reaching way back. I remember the day when the data center people came over to the new client server group and said, yeah, that's not a departmental computer.

We need to put that in the data center. Right? I mean that's kind of what that reminds Me.

Does go back to that. I Mean, we've seen this Story over and over well, but some of that going on here too, like what you were saying in the platform engineering was even the AI itself, I mean, most of our laptops have GPUs in them, right? They're not, you know, these monster GPUs that we're seeing in the cloud AI kind of things.

But, you know, where where does that AI computation, if you will take place? Do we need to be connected back to the cloud or is it on our laptops? Yeah, most of the heavy, you know, GPU AI stuff can't be done locally.

Right. On any sizable amount. You've gotta do it.

Yep. In some farm. So on a previous show, I think it was, you was talking about, we were talking about whether an AI model is an artifact that fits in the pipeline and can you manage it like any other artifact or is it on some orthogonal track that goes alongside, you've been at the show, still got the same opinion?

Or is that evolving? I think it's, you know, traditional physics and quantum computing. You know, they're kind of like the AI never has in the state until you observe it.

And at least it, our understanding of how it fits into the workflow is still pretty ephemeral. And I think a lot of that is just knowing, well, we know how to manage models, you know, for expert systems, et cetera, but training LLMs and protecting data, there's so many things we're learning. Maybe the big enterprises have been doing it long enough.

I, I think for us, I think of it as the common factor is it's the data management problem. It's about the data for the model. The model is just a construct of how it uses that data.

So we can pick up maybe some things from the database world and and from data management platforms, but I think there's a lot to be done to figure that out. That's just my opinion. I can't make up my mind as to whether or not AI will force a restructuring of the way it or teams are organized.

Because today we have developers and data engineers and the DBAs and the, then we got the data scientists and the security people and there are those, um, separation of concerns that we just need to figure out how to work more closely together. Or do we need to throw the whole thing up again and kind of figure out A whole new model? I think you know the answer to that.

They're gonna form new groups. 'cause we, the manager's gotta have a title two that's got AI in it. Right?

So that has nothing to do with the problem. Yeah, it's not the problem we're talking about. No, this is not problem.

That's about the resume, it's culture. Um, um, boy, we sure dipped into sarcasm there. My, my, my fault.

But, but you know, raise your hand if you're surprised that AI's gonna have a big role here, a big presence. You know, I don't see any hands out there in TV land. Um, anyway, moving on from ai, let's talk about our next topic right after this.

All right. And we're back. And Jim Lin, who runs the, um, Lenox Foundation, showed up at a press conference and decided he wanted to kick the can pretty hard.

So he said that the Lenox Foundation is looking into some way to create a, uh, open data foundation, which would make sure that there was always reliable, interesting open data for AI folks to train their models on. Because in the wake of the disclosures around open AI and all these lawsuits, everybody's putting their data behind some sort of licensing agreement. And now, we'll, we won't be able to innovate 'cause we won't have access to enough data.

Of course, you know, he said he wasn't against people licensing data, but he didn't exactly say what you would get if you gave up your data and threw him into this foundation or if there was any kinda prize at the end of the day. But I know we've talked about this in some other shows, but what's your take? So let me be very clear what he's really saying.

Okay, ripping out the, the wholesome goodness of open source school or whatever. What he's really saying is, Hey, we can't train good eye without a good ai, without using your data. And damn, how dare you will make us want to pay for your data.

And so we're looking for a bunch of schnucks who's going to give us, who are gonna give us their data so we could train our AI so that people using said ais can then use your data to make their data and their products better. I don't buy it. I I think that's one time it works when you Anonymize move, right?

I mean, what the only time that kind of works is when you anonymize data to share it with other people. I mean, because I mean, almost by its very nature you're saying, yeah, no, you can take my bid data. It's not really valuable.

Yeah. So what am I giving you? The drugs of data You're giving me government data.

Government data. Exactly. And even the government's not gonna want you to do that.

Now, if this open data consortium or foundation or whatever would somehow make it worth my while that I can create my own LLM not just on my data, but all the other people's open data, and someone who doesn't contribute data cannot use that open data foundation to make their LLM better, well then maybe there's a a reason for me to do that. Well, part of me says try to put something together. If it's valuable, it'll, it'll grow legs and become something.

And if I think, but you gotta have that value prop. You Gotta start Figuring out what is, why, why would I give you my data? What am I getting out of that?

I, I confess I'm having something of an existential crisis about all this. Because, you know, as a person who generally writes for a living or creates video, I feel like I'm on the cusp of creating content for machines, which in turn create more content. And then, uh, you know, I'm basically writing for machines as much, maybe more so than humans.

Humans, it's like training the outsourcers to take your job, Right? Yeah. But, but, but here's the kicker too.

Those machines are then creating content that humans will consume. And you don't get any credit or comp for that. Now there I have read some articles about, I guess in the news world, some of the bigger Aggregators, I'm making deals with open AI Licensing deals and now they're thinking for ai.

So I think that open think what's gonna happen. That seems to me there are a logical, I I mean, look, look at graphics for instance, right? Whether you have, you know, there's any number of graphics clearing houses, it's pretty well settled today.

If you're running a reputable site, you're not just, you know, cut and pasting off the web. Other people's preface. I think we need to have a similar come to Jesus moment Yeah.

Around people's data And For Ai and Of course this, this, this lawsuit's at the bottom of this where, uh, folks are talking about fair use versus, uh, copyright. And you know, a lot of the open AI folks in particular trying to say that, well, we just scraped your websites and it's all fair use, so it's all good. But I'm like, you know, if you're gonna scrape my website to do that, I will put everything behind a reg wall and I will never create any more content to train your AI model.

'cause ultimately we have families to feed and stealing is morally wrong. I don't care if it's legally right, but it's morally wrong. Absolutely.

Let's draw a parallel. That doesn't work. And that's open telemetry.

Is it the data that they're sharing? It's the format of the data. It's, it's getting the data.

That's been an extremely successful open source project. But it's, you know, it's what you do with what you do with the data that you collect, not submitting your data to some open source. Like I said, you, it's gotta be a good value prop for the people who submit their data.

A Lot of work to figure that out. I agree. Agreed.

So Other than that, it'll be successful. I'm fine. Well, we'll see.

Let's, I'm follow calling my union rep. Do you have a union rep? No.

Oh, Good. We need form one. Clearly Just what I need.

I don't have enough on my plate. Your card. I have my card.

Um, anything else you want to add to the discussion? No, this has only been day one. And you can imagine we've had a lot to talk about and on the first day.

So I can't wait to see what tomorrow. We'll be a Lot smarter by. And remember, we are streaming live, it is on Paris time, so if you happen to be in Europe or Asia, it works for you.

Uh, but then we're replaying that stream later in the day for our friends back home in North America or South America. So you can, if you can't make it out here to Paris Toon, you could follow along with us here on Techron. And, uh, it's gonna be a great week.

For Mitchell and Mike and myself, it's a wrap. com is the leading resource for news analysis and education on challenges facing the cybersecurity industry. com covers all aspects of cybersecurity, including data security, DevSecOps, cloud security, application security, network security, security threats and more.

com has the largest selection of security content featuring breaking news, blog posts, podcasts, and more. com to learn more. com.

Home of security bloggers network.