Episode 67: Document Database FerretDB with Peter Farkas, Co-Founder/CEO

Intro

Mike: Hello and welcome to Open Source Underdogs. I’m your host Mike Schwartz, founder of Gluu, and this is episode 67 with Peter Farkas, Co-founder and CEO of FerretDB.

Why FerretDB and not PigeonDB? That’s a very good question. Whereas pigeons are underappreciated for their speed and resiliency, ferrets are fierce and furry and loved by everyone, except Rudy Giuliani.

Mike: A MongoDB API, with PostgreSQL storage, Ferret is a database platform that seems to have a similarly wide appeal. I guess, the 800-pound gorilla in this market is MongoDB itself, with around $35 billion in equity market cap. But in addition to MongoDB, Amazon has also implemented a Mongo conformant API that uses some kind of Amazon storage.
Is there a room in this very competitive and well-capitalized Mongo Cloud database ecosystem for a scrappy start-up with a good technical idea and a track record of reliable operation?

I recorded this episode at the State of Open Conference, or SoCon, which is the largest FOSDEM fringe event. If you can make it to SoCon, I highly recommend it. Unlike FOSDEM, it’s a more traditional event with badges and keynotes, and registration, and coffee, or tea, if you’re British, and after the chaos of FOSDEM, it’s a nice change.

So, this year, SoCon rented around five media vans specifically to record podcasts. So, thanks a lot to conference organizers, it was a pretty great idea. And there’s a picture of Peter and I, donning the Underdog’s headsets on the episode website. Well, with that unusually long intro, here’s the interview.

Origin

Mike: FerretDB was founded a little more than two years ago, you were probably thinking about it for a while before that. What was the spark that led you to actually take action and start the thing, and was the plan always to start a business around FerretDB?

Peter: MongoDB decided to leave its open-source roots around 2018, and all of the co-founders of FerretDB have open-source databases for close to a decade. Peter Zaitsev, one of our co-founders, maybe more, maybe 20 years. After MongoDB went public with SSPL and their departure from open source, we’ve been waiting for a couple of years whether there will be a fork, or if the community is going to do something about it, and nothing really happened.

There was no Fork, there was no OpenTofu reaction to MongoDB’s move. And in 2021, we’ve just decided that something needs to be done. That was the time when we started FerretDB as a side project mid-2021.

Project Launch

Mike: FerretDB has over 8K stars on GitHub, which is fantastic, that’s a lot. How did you go about promoting the project to the developer community to build that kind of support?

Peter: That was the beauty of FerretDB. When we started the project, we were not sure whether someone is going to be interested at all. Since there was no OpenTofu like reaction to MongoDB’s departure from open source, we were not sure whether the community even cares.

If you take MongoDB, there’s not a lot of community around the product itself, in a sense that most of the community consists of users rather than developers of the MongoDB codebase. So, we were unsure what will be the reception. And when we started working on FerretDB, again as a side project, we just created a tech demo, we published it on GitHub, and I think, the first 4K stars came in around two days.

So, we did not do promotion, and we did not expect this kind of outpouring interest in FerretDB. And I think part of the reason why the interest was high is because we’re building on Postgres, and the Postgres community really feels strongly about open source and about open-source databases, and they want Postgres to succeed in various different use-cases, such as document database or MongoDB-compatible use-cases. So, I think that most of the support came from the Postgres community initially.

Product Roadmap

Mike: So, what product features do you see as the most strategic to develop this year, or in the near term? And how do you think these features will help the business?

Peter: As much as it sounds crazy from someone who leads a VC-funded start-up, we are not really concentrating on the business part, because we know that the business success comes after strong adoption and ways to utilize FerretDB in meaningful larger use-cases. So, the most important to us is to increase compatibility with MongoDB. We want the user experience to be seamless, in a sense that if you use a certain tool, or framework with MongoDB, you need to be able to use that with FerretDB as well, without modifying anything in your code.

That’s our goal. But this is definitely not easy. There are a lot of features in MongoDB. Some of these features are not used by many, but would still be used by a very popular framework, and then we need to pay attention to that exact framework to make sure that, for example, Meteor utilizes Oplog, which is a duplicated feature in MongoDB itself.

We need to be very conscious about developing features, which would result in many developers being able to use FerretDB in place of MongoDB, through their favorite framework, through their favorite tools. What’s more important for us is to work with the community on establishing an open standard around document databases.

So, we don’t want to chase MongoDB with their features for eternity, that’s impossible. What we want to do is, we want to work with the existing MongoDB alternatives to establish an open standard, similarly to how SQL came to be, where, if you use SQL, you can query a Postgres database, you can query MySQL database, and any other relational database out there.

And when it comes to MongoDB, there’s no such standard – the best you can do is to become MongoDB compatible. But, it’s like saying that MySQL is IBM DB2-compatible. Do we say that ever? No. It supports the SQL standard, implemented it, and that’s why you can use SQL across the board, with relational databases.

So, we want to reach the same situation with document databases, where it’s no longer MongoDB-compatible, it’s compatible with an open standard.

Value Proposition

Mike: Do you actually have customers who’ve raised their hand and said, “I want to pay you guys for your services?”

Peter: Yeah. We were very lucky. I think, right at the second month, we were bringing in revenue, and that was reassuring feeling that this is something which is sustainable.

Mike: What was the value proposition for those customers?

Peter: I think that the value proposition here is that many customers have existing Postgres infrastructure, and they also have MongoDB. And they need to maintain the internal know-how about these very different databases, they need to pay services for both. And they can just combine the two databases into one and care only about Postgres, run FerretDB on top of it, and they end up with a much simpler infrastructure.

Channels

Mike: Is the open source really your only channel? Do you do any other marketing?

Peter: We are, I think, 80% engineering, and then, we pay attention to channel partners – those Postgres providers who are interested in providing further services to their customers. We don’t run an expensive marketing machine. And we feel that we don’t need to, at this point. We have enough on our plate, on our road map to care about.

Priority of FOSS v. Commercial

Mike: It looks like a relatively small team right now. How much of your time does your team spend on R&D, versus providing services to customers? Because there’s always some friction when you’re providing services – are you working on the product or are you helping customers. And those things can collide. How much time do you spend actually on R&D?

Peter: Initially, we signed a contract, which took away some of our focus from R&D itself, because that customer wanted us to develop features which were not on our road map. And right now, what we are focusing on is to team up with customers who need services and features in close connection to what we have on our roadmap.
Meaning that it’s more like a reprioritization of R&D, instead of going entirely different direction with our development efforts. I believe that with this approach we were able to – this is a ballpark number – but I think 80-85% of our time is still spent on developing FerretDB. Sometimes in relation with what our customers would also want us to do.

Community Contributions

Mike: Have you seen any material code contributions from the community?

Peter: Yes. We had a lot of interesting surprises like that. First of all, there are a lot of individual contributors. And we really like them, because they also want FerretDB to succeed, and they are invested. We have four or five new contributors every month.

The most surprising were the likes of SAP, for example. One day, we just woke up to SAP pushing code in our repo, and that was great. Because that meant that it’s not only FerretDB incorporated developing FerretDB, but