Main content

MariaDB CEO on the open source enterprise - and why good databases are NOT a commodity

Jon Reed Profile picture for user jreed April 13, 2017
MariaDB's first annual user conference in New York City found MariaDB CEO Michael Howard in a confident mood. I decided to push issues, like whether "the revenge of relational databases" favors the incumbents, and see if I could find any cracks. I didn't get those, but I got some spicy/illuminating responses. I also learned why MariaDB thinks its "open source mandate" will carry the day.

MariaDB CEO Michael Howard

MariaDB CEO Michael Howard took the keynote stage at Maria's first user conference (M|17) in a confident mood, and who could blame him? Keep in mind this is the same Howard who said, upon joining MariaDB in 2016, that "This is my manifest destiny."

MariaDB, forked by MySQL founder Michael "Monty" Widenius in 2009 due to concerns about Oracle's acquisition of Sun, seems to be a right place/right time company.

The embrace of open source at the enterprise level, along with the so-called "revenge" of relational databases - all that plays to MariaDB's strength. Not to mention the obsession with the "data-driven enterprise" and the corporate urgency to use data effectively. Digital-or-bust is not a bad market for upstart databases.

Add this to the mix: cloud is not as disruptive to MariaDB as it is to some vendors (virtually all major cloud providers offer MariaDB to customers). Toss in another 2016 funding round, and you can see why MariaDB threw its first-ever user conference this week in New York City.

MariaDB even took the questionable/bold step of inviting media types like yours truly to their first show. So, along with 500 attendees, I was able to ask my pesky questions to MariaDB executives. The question that got the sharpest responses? "Aren't good databases becoming a commodity?"

"The open source mandate is not something that's coerced"

Before I get to that, a word on the open source enterprise. When we started diginomica, we had an eye on open source and NoSQL, but the use cases just weren't proven yet. That's all changed (see Den Howlett's 2015 post, The NoSQL and Hadoop disruptive open source dividend, for more on how our views have shifted. You can't write about digital transformation without reckoning with database disruption).

That theme was central to Howard's keynote, though in a relational DB context. After the keynote, I asked him about the "open source mandate" he was advocating. Howard:

If you were in the database marketplace 15, 20 years ago you, would have said that that [database innovation] was over. Now there's this renaissance... This open source mandate is not something that's coerced. It's something that's embraced. If I were to do an edit on my mandate, it is a mandate of your own self being creative, of trying to deal with the challenges and new opportunities we face in a different way.

The closing keynote had a different theme - the revenge of the relational database. But isn't that a dangerous situation for MariaDB? After all, if the pendulum swings back towards relational, doesn't that play to the strength of some relational behemoths with very deep pockets? No surprise - Howard doesn't see it that way.

Reality check for NoSQL - the revenge of the relational database?

Howard credits NoSQL vendors like MongoDB for driving a surge of adoption and experimentation, but now he sees a NoSQL reality check. Howard whisked me through an oral history of database disruptions, "body blows" to the incumbents - particularly Oracle, the red elephant in the room. NoSQL was one of those disruptions:

When Mongo came out, all these application developers spread like wildfire, right? Suddenly people were saying, "Hey, the relational model is not the best for getting a simple application out." That was a body blow to the relational. Various other NoSQL and NewSQL vendors came out. All of these - in an accumulated way - started breaking down the technical barriers.

So why does Howard think the pendulum is swinging back to relational?

There wasn't enough maturation from a rubber-meets-the-road kind of thing, whether it be Cassandra, Hadoop, Mongo. All of them were part of the time when customers of different sizes were testing out the limits. It's like an incumbent president versus someone new. You knew Oracle was the incumbent president - and you knew all the bad things about it. But these new players - they had to go through their four to eight years to manifest their limitations.

But if the momentum is shifting back to relational, doesn't that favor the incumbents? No, says Howard. He doesn't see Oracle being able to "deconstruct" their business, or going the free/open source route. Nor does he think older relational vendors like Oracle can modernize their products sufficiently: "That's just not the way the world is. They can't suddenly become, for example, a multi-threaded architecture."

For Howard, that all adds up to great timing for MariaDB:

With the price points of open source, it's a perfect storm for MariaDB. Had we come into our own three years ago, we would have been overlooked as a potential heir apparent, because the NoSQL vendors were doing their thing.

Aren't good databases becoming a commodity?

I asked two MariaDB executives my most-likely-to-offend question. We used to say databases will become a commodity. Aren't we heading towards a time when really good/affordable/modern databases become a commodity also? And doesn't that create a value-add problem for a database-focused vendor? CPO Roger Bodamer didn't mince words:

I completely disagree with you. I don't think good databases will become a commodity... It is relatively easy to build a database that excels in a very small niche. We've seen this in things like OLAP, where there's databases that do certain types of queries, which may be ten times faster than others. It is incredibly hard to build a general purpose relational database that works on a variety of platforms, on the variety of use cases that we support.

In the last twenty years, there is no commodity of that stuff. For the commercial vendors, you have Oracle, and I'm not even sure if Microsoft or DB2 have reached that level. Sybase? I don't know. Oracle for sure. In open source, there's only two - MariaDB and Postgres...  It takes a couple of decades to build that stuff.

Howard had a similar response:

Well, that's what we said twenty years ago. If you told me that I was going to be doing a keynote of an open source database company that was going to be the central figure in some of the largest telecomms and banks in the world. I would have said, "You're smoking dope and there's no way I'm ever doing that." There's too many other factors that give me both security and potential.

Howard reeled off some of those factors, which include MariaDB's "ubiquity" on cloud platforms and its Linux distributions ("Oracle even distributes MariaDB because we are a standard now in those distros").

My take

I like to hear a vendor oozing confidence in what they can do for customers. However, I've seen many vendors bruise themselves on the Oracle-can't-keep-up-with-our-innovation rocks. The good news: digital business is so reliant on data - and high performance databases - that this market will surely support multiple winners. Vendor wars are really a distraction. Different databases - different use cases. One ring (or DB) won't rule them all.

Despite the firm rebuttals I received from MariaDB execs in New York City, I still believe the database-as-commodity issue is a danger. There are ways of confronting that - via expansion of services or brilliant tech/platform advancements. MariaDB is pursuing the latter angle, one example being their ColumnStore offering (I have a customer use case on that coming).

It was nice to attend a show where customer value was understood, not chased via a frenzy of extravagant keynotes, slapstick celebrities, and gaudy after-hours events. MariaDB keynotes were brisk and customer focused. To that end, I have a few more customer stories to share. DBS, the largest bank in Singapore, is up next. Let's get to those use cases, and see what they tell us.

A grey colored placeholder image