For MariaDB, it’s time to put the pieces together

As the data platform that began life as the offshoot of MySQL, MariaDB's journey since has been to transcend its roots on a journey to becoming an enterprise-grade database. It's the open source analogue to the journey taken by Microsoft SQL Server, which had similar humbler origins. Over the years, MariaDB has taken on the challenge by creating or acquiring new technologies to the point where, for a small company, its product portfolio began to get complex.

This year it's now tackling that issue, as it prepares for a platform that addresses both small and large scale-out transaction deployments, analytics, and multi-cloud deployment.

MariaDB CEO Michael Howard laid out the agenda while keynoting the company's annual OpenWorks conference being held this week in New York. Steven J. Vaughan-Nichols reported on the upgrades to the security and support features for its commercial MariaDB Enterprise product. That's the part that enables MariaDB to get its foot in the door vying for serious deployments.

But under the hood, the company is now entering a consolidation phase as it starts weaving the pieces into a mix and match whole. That began with the merging of the row-based TX server with the analytic-based AX columnar database for the new X3 platform generation announced back in January. The glue came from its MaxScale database proxy, that was in turn married to a change-data-capture engine working in the background to replicate transaction updates in the row store to column side, and a new query router that would direct queries to the row or column store based on query complexity.

But there are still more pieces that need to fall in place. Last year, MariaDB announced its first foray for offering a managed database service in the cloud with the MariaDB Managed Service, a white-glove service akin to traditional hosted services. During his keynote, Howard pointed to superior benchmarks compared to Amazon's RDS service for MariaDB.

The managed cloud service is just the first step, however. With the more traditional hosted model, it is not the type of service that can economically compete with the self-service managed offerings that each of the major cloud providers offer. Clearly, cloud platform providers have home court advantages when hosting open source databases like MariaDB on their own turf, as they can optimize for their cloud infrastructure. In his keynote, Howard struck a similar refrain that other open source database providers, from MongoDB to Redis and others, have been making about cloud providers profiting off their creations.

For the MariaDBs of the world, the best defense is a good multi-cloud offense, providing enterprises an alternative that gives them the best of both worlds: the economies of scale and flexibility of cloud-native deployment without reliance on any one cloud. At the event, MariaDB unveiled its cloud-native strategy SkySQL, which will wrap MariaDB inside a Docker container whose lifecycle would be managed by Kubernetes. While this doesn't deconstruct the database per se into containers or microservices, it makes managing housekeeping tasks such as deployment, configuration, backups, failover, and the scaling up and down of instances, through a single API, with physical operations orchestrated by Kubernetes.

SkySQL, which was ironically, the original name of the company, will be made available in several forms: for use in private clouds by enterprises that set up their own Kubernetes environments and in the Kubernetes services of public clouds. While MariaDB ultimately wants to craft SkySQL as the portable, cloud-native deployment of its database, initially it will run only in the Google Cloud Kubernetes Engine (GKE) service, as it is currently the most mature managed Kubernetes offering available.

As part of its cloud strategy, MySQL also wants to harness machine learning to make the database more self-driving, which coincides with one of our predictions for this year. At this point, MySQL far away from getting to Oracle's level of autonomous database, as it is just launching an effort to build predictive models, and is reaching out to its installed base to send basic operating parameters, preferably collected over a continuous period at 5-second intervals, to populate the models.

While we're talking cloud, the other side of the coin is what MariaDB will do with its latest acquisition: Clustrix. To recap, Clustrix is a scale-out relational OLTP database that takes its own approach to sharding (they don't call it that): it automatically distributes data but makes that transparent, so you don't have to change the application. For now, Clustrix is a separate product, but MariaDB's roadmap is to incorporate it as the third swap-out storage engine. Clustrix would be MariaDB's answer to Google Cloud Spanner and Cockroach DB, joining the original InnoDB engine of MySQL and MyRocks, aimed at web-scale applications (based on technology first developed at Facebook). We expect that this will happen later this year.

Did we say cloud? As it turns out, almost all of Clustrix's installed base runs in the public cloud. That would make it a logical candidate to leverage MariaDB's SkySQL ambitions.