Exploring KubeCon: Insights from Kodiak’s Co-Founder Ben Ghazi | KubeCon SLC 2024

This is Textron tv. Hey, everyone. We're back here live at CubeCon.

Really happy. It's a crazy cube con that I, you know, I'm always amazed I've been going to CubeCon since, uh, 20 18, 17, something like that. And I always forget the energy at these.

Oh, it's crazy. It has, it's the high, one of the highest energy shows that I've go to year in and year out. So good to be here.

Our next guest is Benghazi. Ben is the co-founder, CEO of Kodiak. That's right.

Kodiak with a C. Ben, welcome. Two Cs.

Well see in, see, see in the front C in the Back. Code maniacs, not the bear Code. Maniacs.

Yeah. Oh, that's a K. But anyway, Ben, welcome Ben.

It's great to have you here. Thanks you for having us. Appreciate it.

So, You know, Ben, look, most of the people watching this live are not here, right? Mm-Hmm. So this is what we want to tell 'em a little bit about, kind of what, what's going on here, but also people always wanna know who's talking, right?

Right. Here's a guy, he's a co-founder, CEO. Everybody thinks they could do a better job than their boss, and, uh, and they have visions of maybe starting a company.

What, what, what about, you know, tell us a little bit about your journey that led you to start A company. Yeah. So Kodiak, uh, I'm, I'm co-founder with my co-founder.

We started it from the very beginning. Uh, we worked together a lot before we started Kodiak as consultants, we're, we're system architects. We built a lot of teams, built a lot of software, some pretty cool stuff in oil and gas, some pretty cool stuff in, in a, a few small startups that we were doing some pretty wild things with machine learning and computer vision.

Sure. Anyways, we kept building these teams and the software and, and kept snagging our, our shirts and our, you know, our teams on, um, being able to ship products, being able to focus on, on software and not the cloud, you know, the product and not Kubernetes. And so we, we sort of had our light bulb moment, um, about five years ago, and we were like, okay, look, Kubernetes is, you know, arguably the best place on the planet to scale up an app and scale up your product and, and, you know, have a product run successfully in the cloud.

Right. Sort of why it was built, but it's sort of like the worst place on the planet for your team to scale up. Right?

It's, it's entrenched in nuance and esoteric stuff, and it's really, really difficult. So we wanted to build a platform that used Kubernetes in all its glory. So you could auto scale things and you could, you know, truly have this really nice system, um, that will scale to the ends of the world, but also have, uh, a usability that your mom could even do this.

Right? So, like, I wanna have a big tent when it comes to building software. I want my QA people in there.

I want my product owner in there. I want my CTO in there. I want everybody clumped together so that we can all see how our software is getting built and deployed, um, so that we can actually have a lifecycle that works for my team.

Might be different for your team. So we set out to fix that. Yep.

Right. So it's, uh, 10 years, well, no. com.

Yeah. Yeah. 10 years ago, 11 years ago, 12 years ago.

The whole way we built, you know, this whole concept of what DevOps is and how it works and how teams work with it, was, was new. Fairly new. Yeah.

And not only fairly new, there were a lot of naysayers. Right? There were a lot of people saying this, all this all nonsense, it's not gonna work, can't work.

And over 10 years, we've seen this whole world ecosystem kind of built up around this, right? Mm-Hmm. Tools.

com, here were your DevOps tools. You had Jenkins and you had Jenkins, you had Chef and Puppet and Ansible, and you had Jenkins, right? Yeah.

That was probably 95% of your DevOps usage. Today we look at tool chains and tools. It's no wonder that like, you know, a lot of people in platform engineering say it doesn't work at scale.

There's so many freaking tools. There's so many different steps in this factory, in this supply line, supply chain. We gotta put in guardrails.

Yeah. We've gotta, It's incredibly difficult, right? So, I mean, look at, look at this place.

It's, it's brimming with people trying to figure out, well, Because you have an organization here that has 180 different projects they're managing. Yeah. Right.

There's probably 20,000, 30,000 people here, actually. 15. Yeah.

This is the biggest one. Trying to Find out what's the, what's the best way. Right.

There are, And the answer is, it may not be a best. Right. There isn't.

There's the best way for every, it's the right tool for the Job. I would say this though, right? C ncf they do great at at bringing to bear, you know, agreed.

A new way, right? Yep. But they're projects, right?

And they're, they're open source. They're, they're, they're not, um, commercialized, right? No.

They're not meant to be commercialized. We're all supposed to work on it together. Kodiak is is sitting on top of a bunch of open source projects, right?

And that's great. The problem we have, and the problem that we're trying to solve is this tool chaining has gotten outta hand, right? Like, how many tools do we need now to operate properly?

And then if we stack up all these tools, who's managing it? Right? Right.

Who's taking care of it? Who's nursing it, who's putting it to bed at night? Who's waking up at three in The morning?

That's why you have these things like platform engineering and sre. Exactly. And all of these, and I'm not saying they're superfluous or not necessary.

There's a place for all of it. There is, but what we've done is we've, it's like as the timeline moves forward, the complexity goes like, you know, it just, It was supposed to get easier, Right. Setting up more complexity.

It got way More complex. So let me, let me throw a counter one at you, Ben. Should we just say, screw it.

Wait for AI to solve all this for us. It's all, okay. I got a great answer for this.

Right? It all boils down to trust, doesn't it? Yeah.

I mean, we were actually just having this, this conversation at lunch. Um, AI doesn't have empathy, right? So sort of by definition, they're sociopaths, right?

They don't care if the app burns down in production, um, because they checked the checkbox like they said they were gonna do Mm-Hmm. And they're like, AI, in my opinion, is sort of the worst thing to put as a DevOps manager, right? Because it's not empathetic for my needs, which is keep my business running.

Um, I, I like the mechanical, right? So like the way we have it today, we, we mechanically make things happen. And that's great.

Now is there should be, there be a layer on top of that that, you know, you can talk to a Jarvis and be like, Hey, Jarvis. Yeah, I wish spin up a new environment for me, uhhuh, and you know, put, put My new suit on too. Put yeah.

Put Yeah. Yeah. It, so I like that for AI in DevOps, right?

Sort of a natural, right? Yeah. Like a a an Agent.

Yeah. A buddy, right. Somebody that I can talk to Assistant, But that assistant has to broker out those things to a mach, a mechanical, it can do it this way, it will do it this way.

It will not have opinion Right. On how to deliver. Yes.

Right? And that it's a double-edged sword, right? Yeah.

Because you want an opinion, you want someone a human. I guess looking at it like in that strange, so kodiak's the answer then. Well, we're an opinionated system for sure.

I mean, we, we built it, we built the, what we built, you know, with, you know, sort of the scars on our neck to be of, you know, look, I, I need a few things out of something like what we built, like truly automated, um, not through A-C-I-C-D pipeline, you know, know, I wanna be able to say, are my things repeatable and are my things, um, uniform, right? com once containers became a thing, sh like, we should have everything should have gotten cleaner and simpler, but they didn't. They actually got more complex.

I knew it wasn't gonna get co It's insane. I love containers. I Love, I love Containers, but like, holy, I remember going to CubeCon, not CubeCon, excuse me.

DockerCon, yeah, yeah. In Austin in like 2017 maybe. Yeah, 20, yeah.

16, something like that. And I came back and I told my team, I said, look, I saw this new thing there called Kubernetes that everyone says is going to, you know, rule how we do our containers and darker. Yep.

I said, but they're wrong. It's so frigging hard. No one's ever gonna want to use it.

Well, famous last words, you Know, it, it is, it is famous last words. You, you, some of the human condition though is like wanting to do the hard thing. Yes.

You know, which is kind of crazy to me. Um, but I, I do it too. You know, I'm, we Own do I we owe Do, but you know, the, so going back to the containers and, and the opinionated part of it is like, we wanted it, okay?

So we broke up the monolith as a, as an industry, and we got into containers and we got into, and then we found that, oh, wow, managing 15 of these things is really, really difficult. And so we wanted to say, no, keep, keep your 15, keep your distributed system and distributed architecture, but treat it like a monolith, right? So treat these things as a unit and treat these things, uh, as one system.

And so that's sort of where we separate ourselves from the pack is like, when we talk about automation and repeatability, we mean it at the system level. And we, you know, each container obviously has to be a part of that and, and will help you along the way for each individual one. But like, the real big thing for us is I wanna create an environment on demand of 15 microservices.

How hard can that, how hard should that be? Could, can I create a a, a production environment right next to production and blue-green? How long would that take?

Is it two months, two weeks, two hours or 30 seconds? Right? For us it's 30 seconds.

I can create an exact copy of production and on a different cluster, right? You know, exactly the same things in, in a couple seconds. And so we wanna be able to say, in environments on demand, right?

Infrastructure, on demand, dynamic config configurations, right? So like configuring at deployment time instead of configuring inside of a CSE pipeline. Got it.

These are the tenants that sort of make us really happy. Drive what you guys are doing. Yeah.

I want to pivot for the rest of our time. So Kodiak is all kidding aside. C-O-D-I-A-C.

Yeah. You see it? io io?

Yeah. Hey, what are you doing? If they take the iOS away, Are they thinking about that?

Oh, you didn't hear this story. Uhuh. I don't wanna be the bear of bad news.

I backup. I hope you got a plan B though. I Got a backup up.

I got a plan B. We own f yamo dev. Okay.

So do io right? Is the top level domain for some island country in the Ian, right? And we Rent, we print these domains don't right From them.

It's a big moneymaker for them. io top level domain. It Sounds bad for me.

Go grab AI or TV or something. Yep. But Yeah, io may not be here by next year.

com. So maybe we just move that, That's a whole different business plan. Um, I'm not going to go there anyway.

io That's right. How do you engage what, what, you know, how do people engage? You can get started right on, uh, right online, just, you know, so we love, uh, user feedback.

We love getting into the community. Um, you know, a a system like this is, is sort of built for the, the, the hands-on developers. And so jump in there.

You create a cluster, you add a couple workloads, you add a DNS, you're off to the races. Excellent. And in, in about five minutes, you'll have something deployed that you can start playing with and seeing like, is this something that I wanna be a part of or not?

I love it. Yeah. Hey Ben, thanks for coming on here, Ben, for having me.

Kodiak do IO for now. Don't worry about the lotion thing. Um, we're live here on Text Drug TV at CubeCon.

We'll be back in a little bit. Stay tuned.