Chen Goldberg, Google | Google Cloud Next ’24

[John Furrier]

Welcome and Introductions

Welcome back everyone, Google Next here live, theCube coverage. I’m John Furrier, host of theCube with Rob Strecce, Savannah Peacert and Rebecca Knight is all here. We’re getting all the action. Day two of three days of wall-to-wall coverage. It’s great to see Google next level. We have aCube alumni here, Hen Goldberg, vice president in general matter of Kubernetes and serverless at Google.

Hen Goldberg’s Role at Google

She’s working on all the cool stuff that’s sitting right above all the CPUs, CPUs, Hen, great to see you. Thanks for coming on theCube.

[Chen Goldberg]

Thank you so much for having me. It’s always a pleasure.

[John Furrier]

Kubernetes’ 10th Anniversary

It’s been quite a journey. We were talking before we came on camera real quick that Kubernetes is celebrating its 10th anniversary. We were reminiscing, looking at old footage from 2017Cube interviews. So much has changed, but there’s a lot of similarities between Kubernetes and AI.

AI and Kubernetes Similarities

But first, congratulations. Well, congrats to all of us for Kubernetes.

[Chen Goldberg]

Definitely, it’s an amazing testimony to the community because I think it’s not easy to sustain a project for so long. And I remember those early days thinking about will there be a Kubernetes 2.0 and what will migration look like?

[00:01:08]

And it was really like our north star was to avoid those kind of migration projects for our customers and users. So I think the community has done an amazing job.

[John Furrier]

Yeah, a lot of debates, but a lot of solidarity, a lot of great community came together. Now AI has got a similar kind of view in the sense of you can see the future, you get it, it’s coming together, it’s unfolding right in front of us. And you have a role and you have a team building out on top of Kubernetes, which is everything from containers to server.

Hen Goldberg’s Team and Role

Take a minute to explain what your current role is now. How do you fit into this? I won’t say new Google, but let’s say Google Cloud. I mean, you’ve got a CEO of public sector with board of directors. You have now all this horsepower, starting to see the layers of the stack and the workspaces, user experience. What’s your team do? How do you fit in to the Google Cloud equation?

[Chen Goldberg]

So my team is responsible for what we call modern runtimes, meaning when you want to come, bring to the cloud your workloads, no matter if it’s a traditional workload, AI workload, any modern workload.

[00:02:05]

We are helping our customers to manage and operate that at scale. It probably sounds familiar with Kubernetes. And we have two main offerings that we do it through. One is Cloud Run, think about it as a container, as a service, the easiest way to build applications on Google Cloud. And the other one is when you’re looking for something more customizable, flexible, you want to build your own platform on Kubernetes, GKE got you covered. And within Google Cloud, that’s exactly the role that we are playing. And we are helping our customers run geo-transformation, innovate, create new experience. And now with AI, it’s just amazing to see what’s possible.

[John Furrier]

Keynote Highlights and Innovation

Things are coming together. What’s the keynote highlights that you see that you got excited about that your team’s working on, where you see opportunities for more innovation?

[Chen Goldberg]

So the first thing that I will highlight is the opportunity with AI to improve our technology team’s velocity.

[00:03:08]

So we made one announcement, Gemini Cloud Assist, a few months back, and I’m sure we’re going to talk about it with Gabe. But yesterday we also announced Gemini Cloud Assist, which is talking about how do you take Gemini and create Gemini-powered insights, and recommendation within the console, and just help me be more productive, be more effective. So that’s something that I’m really excited about. I’m asking my team to become more effective, and productive, and really leverage AI. And we are doing it in the context of testing and coding. So I think that’s one area which is super exciting.

[Savannah Peacert]

Cloud Run and Kubernetes Adoption

It would seem that it’s also with Cloud Run and with Kubernetes and getting up and running, a lot of people don’t necessarily want to run their own kit right now. And is that where you’re seeing the biggest uptick from a Cloud Run perspective?

[00:04:02]

[Chen Goldberg]

I think that’s what’s happening, is that five years ago, I think many wanted to build their own platform. But the world is changing. There are more things we want to do. There are more investments we want to make. It’s actually higher to higher great talent. And I talk with different customers. I’ll give some examples. And those VPs, the CIOs that I talk to, they are saying, I want my team to invest where it matters, where we want to innovate, with our unique IP. And I’ll just give two examples. So we had the founder of Lungchain, which is a very successful open source.

[John Furrier]

Paris and Chase, great guy.

[Chen Goldberg]

So they did a benchmarking on what platform would be the best. And they decided, we’re going to go with Cloud Run.

Cloud Run Use Cases

Why? Because it gives us the velocity, it’s the best for us. And we’re like, yay. But you know what? On the flip side of that, I’ll give two other examples.

[00:05:00]

Today I had a session, my spotlight, when we were talking about the new things we are working on. And Farhan Tawar, he’s a VP at Shopify. And we’ve been working very closely together. And about a year ago, I was asking him, hey, what about Cloud Run? And he’s like, yay, we always do our stuff, we build our own. I’m like, what about trying it? And he’s an amazing leader, and his team is amazing. And they tried it out, and now they’re like, yeah, we’re all in.

[John Furrier]

Really? They went from tired, kicking, do you a favor? Give me a taste, taste test.

[Chen Goldberg]

It’s not for me, right? I think it’s for them, where they invest their time. And we see enterprises that, they just feel that value. So, I think that we will see more and more use of managed services. We can see it with Vertex AI, okay, as well. And I love it, because we would like to build platforms and build services, and I love to see what it enables others.

[00:06:01]

[John Furrier]

Why Cloud Run Works Well

Why is Google Run working so well for folks? What’s the reason, simplicity, the scale?

[Chen Goldberg]

Simplicity, oh my God, you see me smile, right? You know, every time I think that we made it the simplest ever, then the team comes with another way to make it even simpler. So this week, we announced a Cloud Run application, Canvas, which you can use natural language, say what you want, and behind the scene, with Powered with Gemini, we are creating the architecture, all the APIs enabled, everything you need. You just click deploy, and voila.

[John Furrier]

Magic.

[Chen Goldberg]

It’s not magic, it’s technology. But it works, and it makes things simple.

[John Furrier]

Well, that AI gives a feeling of magic. I mean, that’s what ChatGPT did to the whole world, when it said, wow, it’s just getting a response to a query, but it was generating a response. That’s what people want, they want ease of use coding.

Impact on New Application Developers

How’s that going to impact new application developers? Either, is it a low-code, no-code? You see it much more about creativity.

[00:07:01]

What’s that going to do for the developer? Obviously, it’s going to create some value. Time.

[Chen Goldberg]

I think it’s, I have lots of ideas around that topic.

Experimentation in AI

But maybe, going back to Kubernetes early days, I think that the most important thing that we are seeing people do right now, they are experimenting. Okay, and there are a lot of areas that are experimenting. They are experimenting in frameworks. They are experimenting in use cases. They are experimenting, like, which model should I use? Okay, we see a lot of options. How do I decide? And I think that in the next, I don’t know how long it will take, probably not much, because everything is moving very fast. We’re going to see a lot of learnings and experiments, which will then tell us what will be that developer experience. What will people expect? And from an innovation perspective, I will say that we see a lot of AI startups that are building their own models, and they are solving problems in very creative ways.

[00:08:07]

Accessibility of AI Technology

The back of my head, I’m thinking about, for this to really change the world, we’ll need to make sure that the technology is accessible. And now maybe it will be the last thing I will tie to Kubernetes. When we started that journey, one of the things that excited me is that, through open source, we made this technology accessible for people. So no matter where I’m from, where in the world, what kind of education, I can try and scale. And we enabled a lot of experiments and learnings. We didn’t try to solve all of the problems at once, slowly. I think that will be key for such a technology.

[Savannah Peacert]

CloudNative Ecosystem and Kubernetes

Yeah, I mean, I think one of the things, and we were over, I was over at KubeCon in Paris, and one of the really interesting things was that it seemed like, it’s called KubeCon, CloudNativeCon, but it looked like the CloudNative part was really becoming more front and center, and it was the ecosystem around that.

[00:09:08]

How does that really impact what you’re building and how your teams work with all of the different ecosystems that’s around Kubernetes in general?

[Chen Goldberg]

Kubernetes’ Success and AI’s Impact

I think that before everything happened with AI, our goal was to keep Kubernetes boring. I think that’s one of the critical success factors in order to get to the 10 years anniversary. And we’ve done that by creating those abstraction and really enabling that ecosystem that you’re talking about that I think is crucial for any innovation. What I think is now happening, though, specifically with AI, and I’m sure you also heard that in Paris, and we’ll hear about it in the next one as well, is that Kubernetes and containers have some key attributes that are a great fit for what’s happening.

[00:10:03]

Containers are great for innovation and moving quickly. We have an abstraction to the infrastructure, so it’s a good way to manage GPUs and TPUs, and of course I have orchestration, which helps me to optimize. But together with that, there is some new things that are needed. The scale is different, okay? We always take pride in our 15,000 nodes cluster, but then we have customers like Character, as an example, that they need a much bigger cluster. Then what? How do we do that? How do we work with them? So I think this will be interesting moving forward.

[John Furrier]

Accessibility, Security, and Compliance

You mentioned accessibility. Obviously, you’ve been dealing with all this and developed all the time. You’ve got performance availability, scalability, maintainability, those are all the usual conversations. But accessibility is good, but it also brings up two other factors that are part of the AI. I call it the glue layer, it’s emerging relatively fast.

[00:11:01]

That’s security and compliance. So governance has become a big topic, because if you get the governance right from day one, a lot of the data can be scaling and be fast and be into AI much better, more safety, more security. So talk about the security compliance that comes along with accessibility. How do you see that unfolding? Or is that too early for the conversation?

[Chen Goldberg]

Security and Compliance in AI

I think it’s a great point. That’s definitely an area that we invest a lot in at Google, at every layer, okay? So my team, of course, is focusing more on the containers and the runtime from that perspective, and the data team. But of course, we care about data in transit. So there’s a lot of factors there and who has access. So that’s one piece. Another piece, especially when we talk about sovereignty, for example, a lot of people are worried like what kind of access we will have as cloud provider for things running on our cloud. So that’s another angle that we are working on.

[00:12:01]

But that, we continue with your data is your data.

[John Furrier]

Container Security and SBOMs

And container security, I mean, the big concept that’s been always there is supply chain. SBOMs have been a solution. Any update there? I mean, I’m just curious, because since you brought it up, I might as well ask.

[Chen Goldberg]

So we are working on that for sure. And just trying to think about a couple of things. From a container security, a lot of things are staying the same. But what is becoming interesting that with MLOps, so our tool chain is changing. Okay, so working through that for sure. So those are the kind of thing that we are investing in right now.

[Savannah Peacert]

Platform Engineering and Skills

Yeah, and it seems like that as the ecosystem grows and there are new personas coming in, like the persona of a platform engineer. And we were kind of talking about that earlier. And how do you look at addressing that? The skills are changing. And platform engineers sometimes have a much broader view.

[00:13:01]

And they’re more than just Kubernetes. How do you see that as you build out the services like Cloud Run and GKE?

[Chen Goldberg]

Platform Engineering and AI Innovation

So first of all, I think this is a super exciting time for a platform engineering team. Because I think at the heart of it, if you’re a platform engineer, you want to enable innovation, and innovation is happening. We see all of our customers are thinking, how can they take AI? How can they empower their developers to build game-changing experiences? So I think that’s very exciting for everyone in the field. Our goal, also with GKE, by the way, is to make things as simple as possible. So if there is a problem or a challenge that we have already solved for you, and we can automate, we’ll do it. I’ll give you a different example. A Ray framework, which is being used by many of AI developers for their workloads. It’s actually pretty complicated to have it available on a GKE cluster or on a Kubernetes cluster.

[00:14:03]

Simplifying AI Workloads on GKE

Now, for us, it’s a checkbox. Okay, we automated that. Why would you invest the time if it’s a problem that we can solve for you? You want to enable TPUs, it’s a checkbox. Other problems that we are solving, specifically for AI workloads, which I think is, again, interesting. One of the challenges, especially in inference, is that the container images are much bigger. Okay, sometimes they have the model, and if you want to scale out quickly, you need to take into consideration the cold start time. So what do you do if you don’t have a solution? Then you over-provision. But GPUs are very expensive. You don’t want to over-provision. So we are creating a lot of new mechanism in the platform to do that. And I think what, for example, we’re doing the image pre-loading, which Vertex AI, which is also running on GKE, we like to drink our own champagne, have seen 29 times improvement because of that.

[00:15:03]

So we are making an effort to solve everything we know at scale.

Platform Team’s Role in AI

And I think if I’m a platform team, my role hasn’t changed. My role is to enable tens and hundreds and thousands of engineers to build new AI applications. And I shouldn’t forget about all the other workloads that are running. Okay, so my job, maybe from that perspective, is becoming more complicated.

[John Furrier]

Team Size and Organization

And so how many, tell me about your team. How big is it? How are you guys organized? Do you go by technology? Do you organize by groups? What’s the, how do you, can you, it might be confidential information.

[Chen Goldberg]

We don’t talk about size, it’s a large team.

[John Furrier]

That’s okay.

[Chen Goldberg]

We are.

[John Furrier]

But you got a lot going on. You got the container, it’s the big part of the tool chain. You mentioned ML Ops. This is going to be the hottest area on the planet because you’re going to have models coming in.

[Chen Goldberg]

Team Structure and Expertise

So the way we are usually structured is that every team thinks about what’s their role in Google Cloud. And I think you asked that in the beginning.

[00:16:01]

Our expertise is in that runtime piece and orchestration and integration to other systems. So we are not building new ML Ops. But we will be enabling, creating new tool chains, as an example. And I believe that we are doing a good job enabling other platform teams. For AI workloads, because just last year we’ve seen growth of 900% in GPUs and TPUs on GKE. 900% one year. So, I guess we’re helping that innovation.

[John Furrier]

North Star Goal and Vision

And your North Star goal, like Kubernetes to make things simpler. What’s your overall goal that you’re trying to achieve, your vision? Is it to have the most cohesive integration layer? Fastest runtime?

[Chen Goldberg]

In essence, I believe that I would love to see customers use more managed services as much as possible because you don’t need to reinvent the wheel.

[00:17:03]

And I want to make sure that if I need more complexity, if I need more flexibility, we are not creating barriers for that customer. So really allowing innovation to move on. So we are really investing in that, what we call interoperability. So we started talking about GKE and Cloud Run, but now we’re also talking about GKE and Vertex. Maybe I will start my work with Vertex and maybe there will be one use case that I will need for inference and I will be really specific about what kind of utilization I want to get. You know what, maybe GKE is great. Can I use them together? Can I use my model garden with it? Yes, I can. Can I use Colab Enterprise Notebooks? Yes, I can. So thinking about that, I think from my perspective, is how we enable innovation.

[John Furrier]

Distributed Computing and Interoperability

I mean, distributed computing is the paradigm. I mean, making things work, it’s going to be a big deal. Final question for me and Rob might have a couple more questions is that, what’s your business plan for the year?

[00:18:00]

Business Plan and Goals for the Year

Obviously, we just came back from KubeCon EU in Paris. We got North America coming up in the fall. In your job, what is your business objectives, goals for the year? Can you share your plans and what you’re trying to do and what you hope to accomplish this year?

[Chen Goldberg]

So I think the good thing that our mission or our vision hasn’t changed. Meaning, for the past few years, and you know, I can also maybe go back to the beginning of Kubernetes, it is about enabling innovation. And maybe there are different challenges and maybe different workloads, but in essence, it’s about making things easy, okay? Comprehensive, okay? You don’t want to stitch things for no reason and that’s something we take a lot of pride in GCP. Can we integrate the entire stack? Okay, I don’t want to build like point solutions. It has to work together and it has to be reliable, okay? We are a cloud provider. You should trust us with the most important workloads. So across that, we continue on that mission.

[00:19:00]

You should expect us to continue and invest in security. It’s a high priority for us. The second thing is supporting AI workloads. And the last thing is, how can I scale it within the organization? I would love to see more enterprise customers, more traditional workloads benefiting from all those new technologies.

[John Furrier]

Commercial and Public Sector Opportunities

And the commercial opportunities are great. Even the public sector, we had the new CEO Cara Diffitt on earlier. She’s going to be, I mean, all this AI is going to, I mean, think about the government, Rob. All those procurements, paperwork, inadequate processes.

[Savannah Peacert]

Distributed Cloud and Sovereign Clouds

Well, I was going to say, I mean, and part of what you have to be concerned with is that Google’s not just running in Google Cloud now. You have the distributed cloud and you have sovereign clouds and you’ve got all these certifications now with the government, with the agencies and stuff. Is that a big, do you see that as just another set of requirements that are coming in? And how do you keep that for being simple as well, I guess?

[00:20:01]

[Chen Goldberg]

Addressing Sovereign Cloud Requirements

So first of all, we have a special team that is focused on that. And their role is doing exactly that. So how can we take all that power, all that amazing thing that are coming from GCP, but make them accessible with those set of constraints?

[Savannah Peacert]

Yeah, yeah, that makes total sense.

Future Goals and Expectations

What is the, my last question would be, what do you want to be able to be on stage with us next year to be able to say? What do you hope to be able to say next year that you can’t say today?

[Chen Goldberg]

I would love to see more and more customers running AI workloads at scale. We see a lot of the startups now innovating. But I expect this will change how we live our lives. So I think that would be amazing for sure. And from a technology perspective, I think it would be interesting to think about what is that AI platform will look like.

[00:21:00]

But you know, the painful truth, we’re not excellent at predicting the future.

[John Furrier]

Conclusion and Thank You

Yes. Well, you’ve done an amazing job. We love you on theCUBE. And again, 10 years ago, Kubernetes started and CNCF wasn’t around then. Then when they picked up the project a few years later, you were a big part of that success and the community’s glad to have you. We’re glad to have you on theCUBE as a CUBE alumni. You’re like a contributor. You’re like an analyst for us. You’re running all the good stuff at Google.

[Chen Goldberg]

I think, yeah, the key part there, from my perspective, was a focus and optimizing for learning and really working with customers. I mean, that was, from my perspective, so something actually pretty new that we’ve done in an open source project, both creating focus and think about like a product, in a sense, and that’s my philosophy for innovation.

[John Furrier]

Generative AI and the Future of Computing

Well, and I think you’re on the right track. I’ve been saying on theCUBE, I’ve been like a broken record. Generative AI is just distributed computing with the notion of runtime assembling things, and this has been around computer science principles, so it’s kind of another shot in the arm for computer science, engineering, and with low-code, no-code tools, the creativity barrier can be right in the front line, so you don’t need to be a super coder.

[00:22:17]

You just have computer science systems thinking, and you can make things happen, so I think the revolution’s legit. It’s very high-bubbly right now, but it’s a good bubble. It’s not a bad bubble, but it’s a good bubble.

Thank You and Outro

It’s fun. It’s fun, it’s a lot of fun. Hed, thank you so much for your time.

[Chen Goldberg]

Thank you so much for having me.

[John Furrier]

Hed Goldberg here on theCUBE, Vice President and General Manager at Google, covering serverless containers at Kubernetes. Here on theCUBE, I’m John Furrier with Rob Strecha. Savannah Peterson and Rebecca Knight. Dream Team is here. Team coverage here at Google Next. We’ll be right back after this short break.