Data Management Should Build Down From the Agent

Hi, everyone. Welcome back to our live coverage of KubeCon here. We're right on the show floor.

If you couldn't tell, we're shooting out there. There's all kinds of people walking around. It's another KubeCon.

Interesting tidbit I heard about this KubeCon is 40% of the people here are first timers. Wow. They've never been to a KubeCon before.

That tells you something about what's happening in the Kube world. Yeah. That's a number to remember.

I want to introduce you to my next guest. His name is Barry Morris, and Barry's with Couchbase. Barry, welcome to Techstrong TV.

Thank you very much. Great to be here. It's my pleasure.

Let's start off with a little bit about you. Tell the audience a bit of your journey. Sure.

A little bit of background about me. I've spent most of my life actually in CEO roles, running database companies, about half a dozen of those. I spent some time running most of the non-relational database services in AWS.

Mm. I was the VP of product management for the Google databases. But I'm now with probably the most exciting project, which is taking an extraordinary technology, namely Couchbase, and taking it into the agentic era, into the world of AI.

You know what? I wake up every day and I try to remind myself that we're lucky to be here at this moment in time because everything is disrupting, special. Things are changing.

It's super interesting. And there's such promise. Super interesting.

It really is. In the database world, there've been these phases from- Yeah ... mainframe and client server and sort of NoSQL and so on.

This is the most interesting one. Yeah. So the- Especially for databases because at some level, whether you're talking about running your own RAG or your own model, or it's all about the data, storing that data in a way that we can get it.

We say this, and you can say that we've got a particular perspective, but models are very exciting and so on, but you kind of rent them. You sort of switch from one to the other, and you use a different one next year, and so on. What doesn't go away is your data.

Data. And the data is where your long-term value is and so on. Yeah.

But the problem is managing data in an agentic world is massively- Very different ... different from the world. Yeah.

Well, it's a question of scale. And it is, as you said, yesterday we had a breakfast seminar with a couple of panels, and one of the things that came out, you talk about models being rentable. Today, you could use Claude or Chat to go build your own model.

Yeah. It's not a big deal to build your model. Yeah.

It'll build the model for you. Yeah. Point it at your data, and you have your own model.

And I really do think that that's the future. Mm-hmm. Why am I paying these crazy tokens?

Yeah. I'm going to build my model based on my data. There's- Control it ...

there's going to be a lot of change in that world and it's not directly our world, but it is driving our world. And so the advent of these models and of AI and particularly of agents, do you know that already, in terms of internet queries, or search queries- Sure ... over 60% are already machine generated.

Generated. More than half. This world is hitting us and the requirements are extraordinary.

It's like 5X, 10X, 100X times more than what a human- Yep ... would do in terms of data management. You know what I find extraordinary though, Barry, and you've lived through this probably too, we sort of b***h and moan about it- ...

but yet it happens, right? Yeah. We can b***h and moan all we want.

It just keeps chugging along. Yeah. Elections, they come, they go, the data centers, this, that.

Yeah. But progress doesn't stop. Yeah.

It doesn't stop. And I was trying to tell people, we were discussing this on Techstrong Gang earlier this week, or last week, with the elections. Look, I'm not here to do politics, but don't think that somehow this ends AI or what we're going through- Oh, yeah ...

every day is another day forward in what we're living in. Yeah. If you don't mind, I want to turn to Couchbase.

Sure, yeah. As I told you before we went on, I was there when Couch met Base, right? It was two great open source projects, right?

Couch, and I forget, what was the Base? Do you remember? Well, there's CouchDB.

CouchDB. Yeah, which was the sort of database side of it. And then the leading caching system at the time, it was called Memcached.

Memcached, right. Yeah. And really that's very key to what Couchbase became.

It's still true to its roots, yeah. Because although it's now an extremely rich and broad data platform, when you look at that DNA, it's there. It's that extreme high performance, extreme low latency, extreme throughput, very high reliability, handling some of the biggest workloads in the world.

And a lot of it comes from those roots. From those roots. Yeah.

And that, I'm going to guess, that had to be around 2010, '11 maybe. Yeah, exactly. Yep.

As I say, I remember it. That was right at the- I remember the interest, yeah ... heart of the kind of the NoSQL revolution- Well, NoSQL was just- ...

or whatever it was ... right. It was a wave that wrapped over.

Yeah. Because remember, I remember doing a podcast back then- The CEO of Couchbase and the CEO of Mongo. Yeah.

And talking about NoSQL and security. Yeah. That's how long ago.

And back to your point about the AI, what NoSQL was all about was kind of systems of engagement- Yeah ... as opposed to systems of record. Systems of record, yeah.

And relational databases, which I love, and I've built relational database companies and things, but relational databases, their origins are back office and not dealing with mobile and web and edge- And edge, exactly ... and all that stuff. Yeah.

But the interesting thing about this agentic era is it's that same thing, it's that system of engagement- Engagement ... but turbocharged. And so that's why we're so well positioned.

It's a system designed for this enormous scale and so forth. Well, that NoSQL... Look, we've seen graph databases and time series databases and all of these different things.

Yeah. But really, it was NoSQL that kind of set the world free- Mm-hmm ... from that kind of- To think very differently.

Yeah. Yeah. Yeah.

And I think you hit it on the head in terms of systems of engagement versus record. Mm-hmm. Yeah.

It's like you want your systems of records. We need systems of records. We can run them on the mainframe if people want to do that.

But it's- Yeah ... it's not where the real action's taking place. Yeah.

Talking about real action, you guys are here at KubeCon, I know. Yes. " So, let's start with Couchbase is inherently a distributed system.

Any given cluster elastically scales in and scales out according to load, but also we can replicate between clusters. And those clusters can be in different countries if you want. Sure.

And so it's very much a sort of a cloud-oriented system, whereas traditional databases, of course, were scale-up systems. You want to go faster, you get a bigger machine. But you've got to be able to manage those clusters.

And how do you do that? Well, it turns out Kubernetes is a great way of doing that. Yeah, it is.

We have a very high quality, level five operator, which means an autonomous operator, arguably the best in the industry, for managing a distributed database from a Kubernetes perspective. And many of our biggest customers, and we've got some very big customers, Amdocs, for example, which is running a lot of the world's telecommunications networks. Everything that they're doing, application down to database, is Kubernetes based and driving through Couchbase, of course, using that.

So we're very much committed to Kubernetes, if anything, investing more in it. But when you think about what we're doing, it's not lightweight stuff. It's very high-end handling- Yeah ...

very high-performance things, which is naturally what Couchbase does. And that's why you're here. Now, just a couple of weeks ago, last month, just a couple of weeks though, you guys were down in my neck of the woods in Fort Lauderdale, and you had your annual user conference.

Yes. Lot of announcements coming out of that. Yes.

Let me- Great, yeah. Share. So let me talk a little bit about that.

So our Catalyst Conference, it's our annual conference for senior IT leaders, and we made a number of announcements. First of all, we announced around what we call our AI data plane. This is a product that we launched in June.

And what it's about is, uniquely, we're taking the view that data management should build down from the agents, not up to the agents. In other words- Excellent ... instead of taking incremental changes to a database system so it's slightly better for agentic systems, why not build a system that's perfect for agents and then figure out how it handles the rest of it?

So that's what the data plane is about. It's a system that's designed to handle all of your data needs, from performance to governance and so on, for agents. And that was announced back in June.

At Catalyst, we announced a number of incremental things for the AI data plane, notably a bigger relationship with Amazon, with integration with their Bedrock and model management- Sure ... and Bedrock agent management. So a lot of things around that.

Lots of things around the AI data plane. And the other thing that we announced is a new strategy, which is called Converge. And Converge is about Couchbase being able to run workloads that were designed for other databases.

And why? Because our customers, the big kind of global 2,000 if you like, many of them would love to rationalize the number of databases that they're running, but they want to do so without having to do much coding. And so Converge for MongoDB, Converge for Redis workloads, both of those were announced in preview.

And the idea is you can literally switch your application to point at Couchbase, and you get all of these benefits of Couchbase beyond what you already had. Think back over your career, honestly. Do you ever think we've come to the point where it's just push a button like that?

It freaks me out. It does. I'm telling you the truth.

It's just like... Because I've been through database migrations. Yeah.

And I don't have to tell you, they used to be painful. Yeah. It was so painful.

It had to be a really good reason- Yeah ... that you were going to- But now, what's going on in the industry is there is a consolidation of vendors. We were talking about 2011.

In 2011, let me guess, there were probably 200 NoSQL vendors. Absolutely. Today, it is a handful.

Oh, yeah. And the APIs are also consolidating. The MongoDB API is becoming, if you like, the general purpose JSON API.

Lots of people want to use that. And we should all recognize that- Absolutely ... and embrace it.

So today we have SQL and the MongoDB API as the general purpose APIs for databases, and lots of customers are sort of consolidating around those. So- That's the beauty of something like CNCF- Mm-hmm ... Linux Foundation, some of the other daughter foundations, is it allows competitors to cooperate.

Yep. Coopetition. It allows the best code to win.

Yep. So everyone's boat gets lifted evenly. Yep.

And then what you do from there- Yeah ... is up to you. Yep.

Right? Very much so. Yep.

And I think that's the more the market is, and it functions well, because I think the end users are the winners. Exactly, and I do think that the time has come for Kubernetes. It's a sort of a second wave for exactly the reason you're saying, which is that I think there was a sort of a hype cycle thing a few years back where there was a lot of excitement around Kubernetes, wasn't embraced by everybody.

We're seeing it being embraced. Well, it was hard as hell, let's be honest. It is.

Yeah. It's true. It was hard.

It's true. But today, we see the analysts are all saying this. We believe it, we see it talking to our customers.

There's a big move to Kubernetes, and the hardest part of that is stateful systems, databases, where we're extremely well positioned. So... Absolutely.

I wanted to mention one other thing in there, though, is that I think part of this rejuvenation, if you will, of Kubernetes, is that the whole AI stuff is built on this stack. Yes. So the AI stack is the cloud-native stack.

Yes. And when you look at so many of the AI apps and AI programs out here, they're running Coop. Yeah.

They're running containers. Yeah. Right?

They may run different databases, but they need, as you say, systems of engagement-type databases that are going to be dynamic enough- Yeah ... for this environment. 100%.

So- Yeah. It is one of the encouraging things about the AI stack, as you call it, is the adoption of standards across the board, and Kubernetes is one of them, along with JSON and some other things, MCP, the various things that are being adopted very broadly, and I think that's really good news for the industry. Agree.

Hey, Barry, we're almost done, but for people who want to get more information on Couchbase, where can we go? com. Easy.

Easy to get there. All right. Great.

Thanks very much. Enjoy the rest of KubeCon. Yeah, good to talk.

Absolutely. Barry Morris here at Couchbase. We're live.

We got more coming at you. Stay tuned. You're watching Techstrong TV.