The era of "big data" is over, according to Confluent CEO Jay Kreps: Everybody's a big data company now.
Confluent rose to prominence as one of the big data pioneers, built around an open-source project called Kafka that was developed by Kreps, Neha Narkhede and Jun Rao while they worked at LinkedIn. Kafka helps users manage large amounts of real-time data as a constantly moving stream, which is a rather different challenge than dealing with piles of data stored in various places.
The trio founded Confluent in 2014, yet another enterprise tech company built around an open-source project over the last decade. Like many of those companies, Confluent offered additional premium features alongside the open-source project and has since expanded to offer its own managed cloud services that help customers implement Kafka's data-streaming philosophies inside their applications.
Business has been good: Confluent has raised $456 million, with the most recent round valuing the company at $4.5 billion. It's widely believed to be on a path toward an IPO, and although the economic forecast is anybody's guess right now, last month the company hired former Google Cloud finance head Steffan Tomlinson, a veteran of big enterprise tech IPOs such as Aruba Networks and Palo Alto Networks, as its chief financial officer.
Confluent believes that demand for real-time data services is only building as more and more companies modernize their tech infrastructure, and that momentum has drawn the attention of big cloud vendors. Like MongoDB, Confluent changed some of the licensing terms around Kafka in 2018 to discourage cloud vendors such as AWS and Microsoft from offering certain features in their own managed services based around the project.
In a recent conversation with Protocol, Kreps discussed a new cloud service introduced Wednesday as part of Confluent's Project Metamorphosis (get it?), the ramifications of the decision to change licensing terms, and why it's so much easier to work with data systems these days.
This interview has been edited for length and clarity.
What is Infinite Retention, the new announcement arriving Wednesday?
There's been this whole trend towards the use of event streaming in companies. This is now becoming incredibly mainstream, where 80% of the Fortune 100 uses Kafka in some way. It's all about harnessing data in real time. And to do that, it's not enough to just have the data that is happening right now, you have to have the history and context and the rest of the state of the world.
So if you're trying to build a system that manages retail inventory in real time, you don't need to know just what sold right now: You need to know how many products are on hand, and what's the rate things are selling at, and then you can decide if you want to order more or not.
The storage of all that information becomes really important to using these event streams appropriately. The announcement we have [today] is taking all the limits off that storage.
This used to be something that people who dealt with Kafka would have to manage themselves, like how much space do I have, and how much data is stored in each little partition of my stream. Now that whole problem goes away, it's just done automatically.
That's part of this overall project that we're doing called Metamorphosis, which is bringing together event streams and Kafka and modern cloud data systems, this ability to get a system as a service and not think about any of the operational or scaling aspects.
It seems like the concept of "big data" isn't really as important anymore, because it's so much more widely used.
I think there's two meanings for big data. One was just having these data-intensive applications, having more scale, that's become just expected [of] everything that you get in the cloud.
And then the kind of formal "Big Data" stack was really like Hadoop, and there was a set of companies around that. I think that scene really is more of a legacy technology now.
So I think because of those two overlapping meanings, that phrase has kind of fallen out of use. Now big data is kind of just like oxygen in the first sense, in that we expect that of all the new modern systems we would adopt [to have] that ability to scale elastically.
When we first started talking about big data, it seemed like everyone was always thinking about the infrastructure shifts of bringing those concepts into a company, and now it's much more about application building.
The way event streaming came around was very much with applications. It was about fraud detection, and inventory management, real-time customer analytics and customer [visibility] stuff.
For a long time, companies had the attitude that there's this huge learning curve and operational burden that comes with it, and that just slows down its progress. I think the cloud has done a huge amount to remove that.
Now with customers that are adopting our cloud products, they can kind of go from zero to 60 in an hour. It's not, you know, months or more. And that's very different; typically a new data system in a company would take a year, a year and a half to really be production reliable and trustworthy. And so the ability to turn that into hours is actually a really big deal.
Open source has obviously been a big topic for the company over the last couple of years. Do you still feel like you needed to make the licensing changes that you did a couple years ago?
We felt like it was a change we put a lot of thought into, and something that we felt was really important to allowing us to really invest in the part of our product that we give away for free. By doing that, we've been able to double down on that: We've seen a ton of traction in our stream-processing capabilities, in our connector ecosystem, all the parts that we were kind of giving away. By making that licensing change, we felt we could continue to do that in a sustainable way.
I think increasingly, the cloud is seen as not just a product but an ecosystem. And I think there's a ton of attention being put into competing as ecosystems. I think all the cloud providers want to have the best of breed companies providing their offering in the best possible way in their environment, and that's part of why people will choose them.
What has Confluent's pandemic experience been like?
There's two sides to it. There's our kind of internal operations and then, of course, we work with customers often in a very hands-on way. Both of those have had to adapt.
We've had a pretty significant contingent of remote people from the start of the company, and I think part of that came out of open-source work, which is always remote, effectively, right?
So the transition that we made internally wasn't that hard. I think here's a ton of just stress in the world right now economically, people suddenly have lost their child care. I mean, there's all kinds of tension. So that's been something that everybody I think has to deal with, but the actual mechanics of working remotely were not a big shift at all.
In terms of how we work with customers, I've been really surprised how well that's translated. I'm doing probably twice as many customer meetings as I was doing before, because I can do a lot more of them over Zoom.
It's not quite as good as getting to meet with people in person, but you can do more of them and you can spend more time and you can be more prepared as a result, so I think overall quality can be even higher. It certainly made me reassess how I would spend my own time and how much of it would be traveling around versus meeting with people remotely.
I don't know if you can provide a better signal of your future strategic options than hiring a high-profile CFO. What are you thinking about the public markets right now?
It was a really critical hire for us. I spent a long time talking to different people, I'd known Steffan for over a year.
Obviously, we want to build a strong business over a long period of time, we're trying to build an independent company. And we need an experienced CFO to do that.
His experience is incredibly relevant to us, both the time he's spent in world class enterprise companies as well as the time at Google really focused on cloud infrastructure, which is certainly a particular type of product that has all kinds of unique characteristics, both in how the usage models work and the cost structure and so on.
What's the missing link in your product strategy?
There's these two big things, right. There's this emerging paradigm called event streaming, that's been what Kafka is all about. But people's experience with Kafka has been a mixture of these high-level capabilities, and then the low-level operations and tuning and managing servers.
Beyond event streaming, there's a second big thing happening in the world, which is the rise of cloud data systems. I think that that's a much bigger, deeper thing than we even realize.
Can you clarify what you mean when you say "the rise of cloud data systems"?
We tend to think about databases or data systems as some chunk of software: You have a database administrator, and you put it on some servers. The way that you get it in an organization is usually you go to some central team and you say, "Hey, next quarter, I'd like to get access to more database instances, and can you put that into your planning and make sure I have that?" That slows everything down: You have to know exactly what you need and what you want.
So beyond just getting servers on demand, it's being able to get these kind[s] of deep data capabilities. I think that's a huge deal. It sounds like a small thing, but I think it's almost as big an impact as the first generation infrastructure in the cloud.
If you look at some of the systems from the cloud providers, whether it's [AWS'] DynamoDB or Cloud Spanner in Google, you have these amazing data systems. Most of these are proprietary systems that only exist in that cloud, and they often don't have much of an ecosystem around them.
We want to take Kafka and that huge ecosystem around it and things that worked on it, and make it a first-class cloud-native data system, make it something that you can consume and use, the way you would those things that were born in the cloud.
