Heroku’s New Cloud Native Buildpacks with Bob Wise at KubeCon Paris 2024

This is Textron tv. Hey everybody. Welcome back.

We are here at Coup Con Cloud Native Con 2024 in Paris, France, talking to more fabulous, fantastic people, including Bob Wise, who's joining us today. CEO with Heroku. Heroku.

Yes. Yes. And, uh, now part of, or been part of Salesforce for a while.

And, um, love, I really want to kind of catch up and see what's happening, but I'd love to tell us your origin story around old Kubernetes and Kubernetes, kind of what's happening. Yeah, For Sure. You're one of the OG folks.

Oh, OGs in Kubernetes. Uh, so I had been working at a telco, a small telco, and we had been trying to build a Edge Cloud solution, and it was not going that well for us. We were having lots of problems with getting our pipelines and effectively our orchestration, and we weren't using containers.

And I had run across, uh, Joe Beta and Brendan Burns, and they were telling me about what they were doing with Kubernetes. And I got really interested in it. And I had, Brendan came and talked to a group in Seattle.

I'm based in the Seattle area, about what they were doing. And it sounded like such a good fit for this problem. We were working on that.

I went and installed it for the first time myself. 0. Uh, and then I got a really interesting job offer to go to work for Samsung Group.

And, uh, the job I had, we didn't call it this to begin with 'cause we didn't even call it cloud native at that point, but eventually we called the job the cloud native CTO. And, uh, the job was figure out where is the world going with all this, with all this tech, because there are so many different options, uh, what's happening with like virtualization versus containerization. And so we did a full assessment of all the options and we ended up making a really big bet on Kubernetes before Kubernetes even went one point.

Oh wow. Because we felt like the direction of the project was, uh, headed that way. Um, and, uh, you know, we decided that one of the big things that we were worried about from Kubernetes, there are two things we were worried about.

One was, um, scalability, like at Samsung scale, big company scale. Is this Kubernetes thing actually gonna be able to, you're anything size, Anything small on Samsung's come, there may be electronics. Yeah, Exactly.

But everything else is big and those a lot of, yeah. And so, uh, so I ended up, uh, helping form the first sig in the Kubernetes project, which was Sig scalability, because what we wanted to do was help the project and push the project along to making sure that it would scale up to the, to what we needed. 0 launch.

I demoed a little Kubernetes cluster live in the middle of the keynote, which, you know, was a little nerve wracking, but it all worked. So talk about having an influence on the industry. Yeah.

I mean, you have to have a big sense of pride about that. I'm super happy about that. Uh, it's great.

I, I haven't been to KubeCon for a few years because of Covid. This is my first time back. I have spent so much time catching up with, uh, all my old, all, all my old friends in the Kubernetes community.

Mm-Hmm. Uh, it's, it's been really great. Um, so the track record I had, like helping Samsung out there, making that early bet meant that when a W West was looking for someone to launch the EKS business, uh, I got the job and I was super happy.

Uh, uh, one of the most exciting open source projects as a managed service on the world's biggest public cloud. How could you say no to a job like that? And it's a perfect progression from what you've been doing?

Yeah. And then, um, Salesforce made a big bet on Kubernetes. Mm-Hmm.

Made a big bet on EKS. So Salesforce was one of my biggest customers when I was, uh, running the EKS team. And, uh, that's what led to, uh, me, to me taking the, uh, Heroku job.

Uh, it did not, uh, it did not hurt my, my, uh, impulse to join the Heroku team that for the whole four years I was at AWS Every team that would come along and meet with me would say, well, we're using Kubernetes. And the thing we're trying to build for our teams as a platform is we're trying to build Heroku for ourselves. Oh.

And I heard Heroku Heroku Heroku over and over and over again the whole time. Why do you think they were trying to build their own? And It's, it's because Kubernetes was built by design and built by crusty operations people for crusty operations people.

Okay. All Right. You said it.

I didn't, Yeah, I mean, a lot of the problems with the problems we had at the, uh, the, at the telco I was working at is, um, you know, people talk about development and then operations, most comp, most companies Samsung included, if you look at the total life cycle of what are people doing for the dev part, and then the deployment and operations bit, it's like 80 20. And so, oh absolutely. If you, uh, if you improve the development bit a lot, you don't get as much a gain as if you even improve operations by some Mm-Hmm.

And so, uh, Kubernetes really addressed the most painful part, but, um, it's still complicated, hard to use. Uh, you know, developers need, still need some help. And so the reason you hear that over and over again is the operations pain has been largely solved, but the development pain maybe not so much.

Okay. Interesting. So we have made progress on the operations side.

Yeah. We've totally made progress on the operations side, immutable deployments, um, you know, the idea of, uh, uh, you know, you have this packaged thing, the container that it's gonna work the same in development and staging and deployment. And, um, the idea that you don't change things in production, this is the mutable concept that you build the new thing, you throw the thing away in production and you replace it completely.

This is such a move forward in terms of, uh, a kind of operational philosophy. I remember when I got introduced to Heroku would've been before you joined, but, um, it, it was really sort of the, um, the, at the time they would call it shadow it, but it was the way of developers getting access to, you know, instead of opening a ticket, right? Yeah.

The, the traditional problems. Uh, but it was just a really easy environment to get up and running very quickly for development teams and even take it to kind of a production state. I dunno.

Still true. Still true. Is that, is that a motto that has really grown and expanded, or have you shifted focus in what you're doing?

Uh, I would say much like the big cloud providers, certainly AWS like AWS kind of gotta start with shadow it as well. Very much the people come along and put the credit card in. Poof.

Uh, but the Heroku business today is very much a big enterprise business. So much the way public cloud is. It's just, uh, we make it, we make it super easy.

Mm-Hmm. So excellent. I mean, We sure for sure do have the folks that come along and put their credit card in Yeah.

Gets easy entry and That's still a thing. Um, but, Uh, but supporting large enterprise is a whole different property. It's a whole d thing.

You talk about operational, scaling, all the things you've Done. Absolutely. And I think this is where the Salesforce acquisition of Heroku has helped Heroku a lot because Salesforce is amazing operation when it comes to, uh, customer trust and support and like doing all the things on a global basis at scale.

And so, um, you sort of put the two together and you've got something magical Now you're do, uh, where, where do you want to take things next? What do you that's kind of on the horizon? Well, uh, for sure we're doing a lot more now with generative ai.

Uh, and so, uh, we have some, well, for example, we supported, uh, we just added PG Vector into the Heroku, uh, offering, maybe I should say. A lot of people think of Heroku as being, let's say, more of a compute service. You know, you have dyno, you have some code easy, like push the code into production, but half of what we do is databases.

Hmm. Um, and, uh, part of the magic is you get all of the pieces that you need to build applications. You get the, you get the Dinos, you get the compute, you get Postgres, you get storage.

Um, uh, we have Redis, we have the pipelines. You basically get everything you need in a, in a single package. Uh, and so the first thing, uh, the first thing in this space that we, that we ship is PG Vector.

You know, vector database is really important for RAG and other things. Absolutely. Yeah.

Um, we have some really exciting things coming through, through the rest of the year to, uh, uh, help folks with the accelerated inference. Um, I think Heroku is, um, uh, in the AI space, it's probably not the place where people are gonna do like big train, new big models, but it is definitely gonna be the place where people run inference and run production, uh, production things. Uh, It's a lot of interest in, you know, small language models.

Yeah. Uh, domain specific, right? Yes, for sure.

Not the, you know, the chat amputees or, you know, whatever the World and whatever I just said is probably incorrect. Was incorrect. Maybe a, you know, a, a week ago or definitely gonna be wrong two months from now.

We Need the statement at the bottom, you know, no prediction of future performance and all That kind of stuff. Yeah, exactly. No, it's, it's, uh, there's really been even tremendous progress in the last few months, even on, uh, on running smaller models.

So maybe I'm wrong, hopefully I'm wrong that there will be a lot more of these models actually built. Um, but I think what, how we're seeing it at the moment is that's part of, that's part of, uh, what you want to be able to easily run, run and package. Um, what do you, sorry, I don't mean to ask this a negative way, but what do you think we're doing not quite right about how we're thinking of adopting AI and using ai?

Is there sort of an opportunity that like, we need to be headed this way now? Uh, that's a, that's a big question. Uh, Let me constrain.

You can go anywhere you want. So let me constrain my answer to the effect on developers. Um, because, uh, I think this thing has been going on a long time and Kubernetes and Heroku have helped development teams cross this chasm.

And it's a really hard chasm to cross. And the chasm is you build some code, it's back to, it's back to deployment and operations. Can you actually deploy that code running into production without operations people touching it?

And, uh, if you look at the really high performance organizations, they tend to have this thing which is they're terrified if human beings touch the code on the way to production Mm-Hmm. Whereas a lot of teams that are less productive are terrified if humans aren't touching it on the way to production. Okay.

And so that's the chasm and, and getting to a point where you're able to do fully automated continuous deployment is super hard. Yeah. Uh, not continuous delivery, but continuous deployment.

Yeah. Like all the way into, into, into production. And I would say probably the company that has, uh, you know, publicly at least, uh, really mastered that would be Facebook.

Um, and one of the reasons that they, they put, they made that big investment is, uh, they're known for being able to have like very junior people come in, build lots of things, and then not get that stuff pushed into production if it doesn't pass test cases. Lots of guardrails, lots of testing infrastructure. Um, and what we're seeing now, I think is the effect of generative AI on coders is it's, it's kind of like everybody just added another 10 junior developers to their team.

And so That's a good way to look at it. Right. And so you still, like, it means, it means that, um, the investment that's required to cross that chasm of testing and continuous deployment becomes even more critical.

And so I think the big effect here is that the, the companies that have made the investment to cross the chasm are gonna get huge benefit because all these junior programmer robots, um, are going to safely help them increase, increase their productivity, and the folks that haven't crossed the chasm are now gonna need more people to keep an eye on all the new robots. So, and so Is the argument worth making then? If you wanna be successful at ai, one of the things you really need to do is get across that chasm.

You have to get across that you're Not gonna be able to harness all those Absolutely. Right. When it comes to software development.

I think that's, I think that's absolutely true. And the, the competitive gap between the folks that have crossed that chasm and not is getting wider and harder and more important across Mm-Hmm. Very cool.

Um, what are you most surprised to see coming back after a few years? Uh, absent or I just, you know, doing other things covid and wise and what are you like, wow, I didn't expect this. You know, one of the things I didn't expect is, uh, you know, back when I was going to KubeCon more often, the big coup con was in the US and the sort of smaller secondary show was in Europe.

I don't think that's where we are now. Like this is, there's a lot of people here and the energy is so amazing and so, uh, I'm kind of thinking that, uh, it's sort of looking to me like maybe, uh, maybe, uh, KKU Con Europe is like the main event now. Kinda Come on us, come on North America.

Exactly. We, we need to get with the program here. Right.

So it's just amazing to see all the, all the energy here and Uh, there's an all in feel kind of to it too here. Yeah, absolutely. Is what I get.

The, Uh, other area here on the show floor that is super busy and great to visit is the CNCF project area. Mm-Hmm. You go to the open, like go and visit with the, all the folks that are putting all this energy into the open source projects that are part of the CNCF also.

Pretty amazing. It is, it is. Well thanks for your contributions to getting us here.

Thank you so much. Which are many and, uh, the great work you're doing with Heroku and Excellent. Thank so much.

Thanks for having me on the show. See you, Mike. It's been fantastic talking with you.

Thanks. Hope you'll come back again. I will be happy to do that.

Awesome. Alright, thanks. Awesome.

So, um, Bob Wise, who is CEO with Heroku and OG of many fundamental and important things that we rely on every day, so big thanks to him and many others who contributed to that. So, as I said, we have fantastic guests. They don't get better than they do here on Techstrong tv, so stay tuned.

We've got other amazing folks. I don't know if they're the OG of, of things as much as Bob, but they are fantastic guests too. So thanks Bob.

Alright, Thank you.