What's Next for Cockroach Labs? A Conversation with CEO Spencer Kimball
We discuss database survivability across clouds and regions and, of course, AI.
Welcome to the Cloud Database Report. I’m John Foley, a long-time tech journalist who went on to work in strategic communications at Oracle, IBM, MongoDB, and now with Method Communications. If you received this newsletter, you’re a subscriber (thank you!) or someone forwarded it to you. Subscriptions are free.
Hello again everyone. It’s been a few weeks since my last post. I traveled to San Francisco to visit my son, a software engineer who lives and works in the city. We found the streets and restaurants to be buzzing with good energy. To paraphrase Mark Twain: Reports of San Francisco’s demise have been grossly exaggerated!
Then I flew to Chicago to visit a farm in northern Illinois to report on the corn and soybean harvest. It’s the latest in a series of articles I’ve written over the past couple of years on precision and digital agriculture.
And then it was back to New York City where I attended Cockroach Labs’ RoachFest 2023. Highlights included the announcement of CockroachDB-as-a-service on Microsoft’s Azure cloud, and presentations by JP Morgan, Santander, Booking.com, Doordash, and Procter & Gamble. I met with Nate Stewart, Cockroach’s Chief Product Officer; Peter Mattis, Co-founder and CTO; and Spencer Kimball, Co-founder and CEO.
In the morning keynote, Kimball talked about CockroachDB use cases including storing/managing data locally for governance or performance purposes; running the database in a multitude of deployment options (on-premises, managed service, hybrid, etc.), and database “survival.”
Kimball wrapped up his presentation by emphasizing the need to ensure that business execs are on board with an IT team’s data-management strategy. “Executive touchpoints are crucial,” he said. “They are critical to driving alignment.”
That insight resonated with me because, in my experience, many database companies are outside of their comfort zone when talking about database strategy with business people. So it was one of the first things I asked Kimball about during our 1:1 interview.
We talked about multi-cloud and multi-region deployments, the challenges and opportunities of working with global customers, data localization (or locality), and much more. Our conversation ended with a blunt, hype-free discussion of AI and the need to support vectors in the database.
“Our perspective on this is not to just rush in and wave an AI flag,” Kimball says. “That’s short term, chasing the puck.”
I learned after our interview that Cockroach Labs has just laid off an unspecified number of employees. Had I known, I would have asked Kimball about that, and how the reduction in workforce would impact the company’s go-forward strategy and ability to execute. In a subsequent written statement, Kimball assured that Cockroach Labs will maintain a commitment to sustained growth and strategic priorities. For the record, I’ve included his full statement on the layoffs following this Q&A.
But first, here’s the transcript of our Oct. 3 conversation, which has been lightly edited for brevity and clarity.
My question: So much of the conversation in the database industry is technical people talking in technical terms to technical audiences. During your keynote, you made reference to “executive touchpoints,” which I think is hugely important but too often lacking. Why did you say that?
Spencer Kimball: It’s the difference in a year and having these epically large and important businesses trusting us with their data. And you realize in order to make those relationships, the vendor/customer relationship, truly work, you don’t just need the buy-in of the chief architect. Ultimately if you want that not just to be initiated, but to be maintained and expand, you need to penetrate more of that organization, to activate the desires of the folks who are running the business units and overseeing the long-term technical direction of company—how they’re going to run their business efficiently, enter new markets, make money, save money, eliminate risk. That’s the thinking of executives, that’s their mandate.
CockroachDB plays into all of those desires. And what we found is that the chief architect level is a lovely place to start, that is the technical conversation. We’ve always been good at that. But as we’ve matured as a business, we realized that to serve our customers better—and it aids in our growth and success as a company—you have to [engage] the folks whose job it is to see that the future of their business is more than just an application succeeding. So some of the key things we’ve seen is that you really unblock progress when it’s been stymied, at that executive level. And ultimately you’re aligning strategies, because if our customers want something that is different from what we’re driving towards, the projects and the relationships can go off the rails. We want to make sure that what they’re after with CockroachDB is what we’re trying to support. That’s when the relationship works best.
It’s not that we’re moving away from those chief architect interactions; those are our bread and butter. But if you want those to thrive, and that relationship with a customer to expand, you need to really be checking in. How’s CockroachDB working for you? If the CIO is happy but the chief architect is not, that’s good to know, and vice versa. If they don’t see you as aligned with their strategy, you’ve got to figure out how to prove that.
My question: How is Cockroach doing in up-leveling those conversations? I wonder if business managers and business leaders get what you’re talking about with this thing called multi-region?
Spencer Kimball: We just had our Customer Advisory Board (CAB) meeting. It is always eye opening to hear our biggest customers, and often folks who are very high up in the decision-making process, having conversations amongst each other. There’s more openness.
In that meeting, there were two themes: one was multi-cloud; one was multi-region. It’s very in line with what what our strategic strengths are, and we are almost uniquely capable in terms of the multi-region aspect, that locality. That really does lend itself to the world’s biggest businesses, which have global customer bases and are in lots of different markets. Multi-cloud is also interesting, because some of our biggest competitors but also partners are the cloud vendors. But where we are differentiated in an obvious fashion, that’s not strictly technical, is the fact that we run everywhere. These big organizations have a hybrid reality and a hybrid future. Even if they were all-in on cloud, they want to make sure they have cloud portability; maybe they’re actively running across different clouds. That is an explicit part of the mandate, the strategy of these customers.
I sat in a meeting with Microsoft CEO Satya Nadella. He was quite aware of the idea of multi-region but he didn’t describe it as multi-region; he described CockroachDB as “regionless,” which is an interesting and perceptive way of looking at what our real value proposition is. Multi-region is not an easy thing to achieve, and that’s where we put a lot of our technical and intellectual investment.
It’s also not easy for customers. You’re talking about a much larger area of concern— data sovereignty, domiciling, network costs, latencies. Satya described it as regionless—so a customer doesn’t think of the world as regions anymore. It is an experience where those kinds of boundaries and the difficulties associated with them have been transcended.
The fact that it’s on the mind of the CEO of Microsoft is telling. If you go to the CIO of a financial services business, which we routinely do, they know what that’s about, because they’re in multiple markets around the globe. They understand for a number of reasons: How do we provide a first class user experience for all users and not have silos everywhere? And they’re being pinged by their legal teams about the liability associated with, Where is the data domiciled? Where is the legal jurisdiction? Where is the PII data?
So having a technical solution that’s in line with [these] concerns is top of mind. With what Cockroach Labs is bringing to market, we have good timing. For the customers that we’re intensely focused on, this is important to their businesses and their strategies.
My question: Who are the customers that you’re most focused on now?
Spencer Kimball: From the very earliest stages, we have had the good fortune to be in front of some customers that we see as vital to our future. JP Morgan, for example, is one of our earliest. They invested their attention and energy in CockroachDB, when we we weren’t yet in production, still an alpha product. That was a sign of how technically astute they are and how inquisitive they are of new technologies. I think that serves them well.
Over our history of nine years, we have served different segments of the market, not just the really big companies. Where we’ve found ourselves more recently is that it’s difficult to build for the entire market. If you’re trying to do it for everyone, you have to kind of do this lowest common denominator. What’s the marketing message? Because it’s different for different audiences.
As we’ve matured, we’ve looked at where’s our success? Where do we have customers who, when there’s a financial downturn, they’re not only still in business, but they’re still very much wedded to the idea of Cockroach. They want to double down their investment is us and not try to cut costs for something that’s as important as their database infrastructure. So, where do we win? We found is that we need to continue investing in that customer success journey with these companies that have the massive potential.
We call them established data leaders. What we mean by that is their business is established, which puts them in the enterprise space, but more specifically, these are companies that use data to differentiate. Data is very important to them; that’s actually part of what gives them an edge in their respective markets. And then we narrow it and say, Where are the ones that right now are looking for what Cockroach is offering? We see it in financial services, tech, retailing, and gaming. Those are the four where we have a very decided focus.
That helps us when we’re talking about CockroachDB’s benefits. When you can couch it in use cases that are related to financial services, that holds up a mirror to a prospect. They can see what you’re talking about and say, I've got that problem or I’m going to have that problem. If you try to make it general, sometimes you find that you’re just using words and everyone else is using words, and it’s not really a message that’s getting through all the noise in the market.
My question: You used the term ‘established data leaders.’ Is this a new customer segment for Cockroach Labs?
Spencer Kimball: We have always been on that path. That’s where the center of gravity is, and they often put a lot of pressure on our roadmap. That’s good, you want to respond to where your business has the most fertile ground.
We welcome startups and individual developers. We needed them to prove our products. And any adoption provides additional paths to reaching a champion at one of these customers. But you have to be careful that you don’t get into this trap of talking in your messaging to everyone that might listen.
Happily, these other products—cloud, dedicated, serverless—are moving into the stages where these ideal customer profiles—companies that are established data leaders that are in these four verticals—are ready to use those products now, and the products are ready for them. So it’s been a virtuous endpoint.
But you can get distracted on your way. So, why did I stress the executive touchpoints? You start to see the world a little differently when you establish yourself successfully in these big accounts and you see the tenor and the future of your business.
My question: I talk to a lot of database startups, and there’s innovation everywhere. It’s not unusual for startups to target the developer teams to get their foot in the door. And then they’ll talk to the senior business people, or try anyway, once they get established. It’s like hyperscale analytics. You probably don’t need that if you’re an SMB. It’s for big, sophisticated companies.
Spencer Kimball: That’s a crucial point, who actually needs it? And it was quite a lesson when we had the [recent] macroeconomic reset, just in terms of if you have a customer that’s aspirational, but they don’t have a global business or massively scaled user base and things like that. We’re always happy to serve them.
On the other hand, these bigger companies need the system; we are an essential part of their modernization strategy. They often have strategic mandates around CockroachDB. They want to use it for as many things as they possibly can, [like] mainframe modernization. That’s great technology with an incredible legacy, but it’s not always right for the use cases anymore. So what’s the successor technology that a company wants to invest in? That’s an area that Cockroach has a lot of growth potential.
My question: You said that at the Customer Advisory Board meeting, they talked about multi-region, multi-cloud. And you gave an example of a customer that’s doing data replication to three clouds—AWS, Azure, and Google Cloud. Is that an emerging scenario?
Spencer Kimball: We don’t know how many customers are doing that because we have self-hosted customers and we don’t always know how they’re configuring. We do know that we’ve got multiple customers and prospects that are interested in it. I’d say it’s still early for that. Let’s say we’re talking about financial services, where I think this kind of deployment really makes a lot of sense. If you’re running retail banking or a payment system—and that’s actually the company that is doing the cross-cloud replication now—that’s such a valuable always-on service that the costs are outweighed by the benefits.
They’re an example of an early adopter. It’s kind of a luxury capability, but those costs can be brought down. We could offer that kind of capability in a serverless footprint. It even has the chance to become part of a regulatory scheme or a recommendation, like a reference architecture. That’s especially true in the UK where they have the Digital Operational Resilience Act. Database infrastructure is the difficult problem to solve; it’s relatively easy to run your stateless application logic across all the clouds.
That provides an incredible story for what happens when there’s a cloud outage, or something worse, and you have to move off a particular cloud service provider. It doesn’t take you 30 days, which is often what a migration strategy looks like. You can do this in 30 seconds. The failover time would be like four and a half seconds on average for CockroachDB. So you can lose a cloud and your app stays up, it might have a little bit of latency. Once a company establishes that, it becomes a recommended way of doing things. While I do think of these as early adopters, we have to make them successful, and their success is going to breed imitation.
My question: Are resilience and recovery the primary use cases for data replication across two or three clouds?
Spencer Kimball: Replication across three clouds is still in the early adoption stage. But disaster recovery and IT resilience are really what Cockroach is offering. It even extends to routine maintenance. DoorDash had a great presentation. They’re a customer that runs many CockroachDB clusters. They are in a stage now where they’re constantly repaving clusters, replacing existing machines with new machines. And they do that during normal business hours. That’s something you can never dream of doing on a more traditional legacy database; that would be a difficult and fraught sort of operation. But because you have IT resilience, you can kick nodes over, take nodes out, add nodes, repave nodes, upgrade the servers, upgrade the database. All those things can happen while you have peak load. It’s a tremendous capability.
That’s why we’re called Cockroach, that resilience. That is probably the bedrock capability. In financial services, that’s the number one thing, the acute pain point that they’re solving for. Sometimes it’s also [data] locality. In big tech especially, it’s scale, but also with resilience. So everyone has a different emphasis on the capabilities we’re bringing to the table. And by the way, enterprises that run it everywhere—across the cloud, cloud portability, whatever it is that they’re specifically focused on—that’s huge. That’s a big differentiator from the clouds that are offering a cloud-only solution.
My question: I want to ask you about AI. Other database companies are talking a lot about it. What is your AI story?
Spencer Kimball: The macro point of view is that AI does look poised, if it’s not already fully active, in terms of creating a lot of new use cases or refactoring of existing use cases. And that ultimately requires more databases.
When crypto was huge, people were asking, are you going to put CockroachDB on the blockchain? That was never something we considered. However, what’s true in any of these crypto companies is they need a system of record to store all the data about their users and their operations. So any new use case, needs a database.
Now, increasingly, these databases are tasked with storing high dimensional vector data; that’s something we’re working to support. Think of that as just another modality in the database, another data type, another kind of index to search on that data type. It’s not rocket science. Geospatial is another example, or full text that we have indexes on. So you have these different data types.
This is just 1,500 dimensional vectors with different indexes that let you find similar vectors. The important thing is that all of that lives, in our case, in a distributed system that has this kind of reliability, resilience, scale, and required security compliances.
Adding another modality into the database is actually rather trivial. All of these companies that are our customers have AI use cases, and they’re using different tools to build those products. But all of them need to use a database to store their system-of-record data, the metadata about their operations. That’s ultimately what we need to be the best at. Yeah, we’ll add vectors. But we’re not trying to shift our focus to, ‘We are AI. We’re an AI company.’
Because the space of operational databases, and particularly relational operational databases, is very large. And it’s the work of decades to build a system that is so general purpose that it can accommodate every vertical and scale in terms of the company. We’re well on that journey. We’ve chosen a very valuable niche, which is companies that have these problems of scale or locality. It’s from there that we’ll expand, and I think every AI use case is one that we can back.
My question: So I heard you say that at some point, you’ll add vector data support to CockroachDB. Right?
Spencer Kimball: That feels inevitable. But I also feel like that will be just the start of what we’re going to need to properly support AI, because we’re in the first chapter of what these data management systems have to store in order to represent meaning. I think that’s what the LLMs are trying to get at—some deeper notion of things. It’s structured data in a sense, but it’s playing in this much higher dimensional space.
That’s just one way. That’s just how LLMs work. We’re already starting to see that LLMs are very good at some things, but have incredible blind spots in others. It’s not AGI [artificial general intelligence], although people are saying it’s the basis of it.
We’ll see where it goes, but my my guess is that we’ll be tasked with other kinds of modalities in order to support AI, because AI feels like it is the engine of incredible economic growth. So there will be many use cases and a lot more advances in terms of what needs to be stored.
We’re in the earliest innings of it. So our perspective on this is not to just rush in and wave an AI flag. That’s short term, chasing the puck. What are we really good at? We’re solving the true business needs of these massive companies. They change, but they change at a rate that is going to make our product relevant for a very long time.
Chasing the vector angle is a good way to get press, to get venture funding, which we have plenty of. We’ll add all that, we’re in the process of doing it. But we’re looking at what our customers are asking for. They will eventually want this. Some are already starting to feel around the edges. But what they really want is a way to upgrade these systems that hold a large part of the world’s GDP in their use cases.
My question: I’m wondering how the emerging world of AI is going to change the big picture of data management. Will there be more data, more distributed data, more automation? You’re the visionary. How do you anticipate things being different in a couple of years?
Spencer Kimball: It’s a good question. The applications of AI seem to be changing at a rate that is hard to keep up with, honestly. But it still hasn’t dramatically impacted how people need to store metadata, or what kind of metadata they need to store. I can see that you certainly can build use cases more quickly with things like co-pilot; it can increase the velocity of new use cases. It doesn’t seem like it’s going to change our mandate dramatically, but I see virtually everything that we interact with changing. I see that as a good thing for a company like Cockroach Labs.
My comment: Everything you do is going to change anyway, whether AI or not.
Spencer Kimball: That’s the way it’s been going. How rapid is that rate of change? It feels like it has taken a sharp turn upward.
EDITOR’S NOTE: Subsequent to the above interview, which took place on Oct. 3, 2023, I learned that Cockroach Labs has laid off an unspecified number of employees. In a written statement, Kimball explained it as follows: “We have made the difficult decision to reduce the size of our workforce. We are deeply sorry for those who have been impacted and are grateful for their service to the company. This restructuring reflects our commitment to sustained growth, directing our talent and capital to our most strategic priorities. We will continue driving our industry forward and are confident in the unparalleled service we deliver to our customers.”
Needless to day, Cockroach Labs isn’t alone in making the tough decision to cut jobs. According to The Information’s “The Briefing” newsletter, LinkedIn, Stack Overflow, and Flexport are among the latest companies to have made similar decisions. “Whatever job security existed in tech appears to have disappeared,” Martin Peers wrote in The Briefing. Like other companies in similar situations, Cockroach Labs will have its work cut out to assuage concerns while it forges ahead with a smaller talent pool.
If you enjoyed this blog post, please share the Cloud Database Report with someone you know. And watch for my next blog post, which will provide an update on database startups.