In a bid to expand its charter beyond the transactional database realm, MariaDB Inc. is extending the capabilities of its open source relational database with online analytic processing (OLAP) operations that work on the same data store.

The company said MariaDB ColumnStore, which will go into beta testing next month, will be the first database engine to enable both transactional and massively parallel analytic workloads to be processed under the same roof.  The capability is made possible by MariaDB’s extensible architecture, which allows for the simultaneous use of task-specific storage engines on top of the relational core.

The company also announced new data streaming capabilities that enable transactions in MariaDB to be replicated in real time to Hadoop or any other data store.

Although not as flexible as big data frameworks like Hadoop, which handle large amounts of unstructured data from multiple sources, MariaDB ColumnStore offers some compelling benefits to users of the core engine, including a common interface, security and ANSI SQL query support. “Our approach is aimed at any company that does not want to pay big for big data,” said Michael “Monty” Widenius, developer of both MariaDB and its MySQL predecessor, in a press release.

MariaDB ColumnStore will be offered under GNU General Public License, Version 2. The company said it chose GPLv2 rather than the more recent GPLv3 because the license is “widely accepted by community. We feel it’s adequate to match community demands,” the company said in a statement.

MariaDB has been on an aggressive campaign to raise awareness since it closed a $12 million funding round and hired a new CEO in January. Created by the developers of MySQL after Oracle acquired that software with its purchase of Sun Microsystems Inc. and changed the licensing terms, MariaDB has replaced MySQL in the open-source LAMP (Linux/Apache/MariaDB/PHP/Python/Perl) stack and is used by a number of Web-scale firms.

The company offers both free community models and enterprise editions of its products, saying that they are a better choice than MySQL for companies that want true open source flexibility. The software has an estimated 12 million users.

The addition of OLAP capabilities addresses the interest many businesses have in gaining the insights of analytical modeling without paying the cost penalties of data warehouses or dealing with Hadoop’s complexity. “You can use commodity hardware to scale to hundreds of terabytes of data,” said Nishant Vyas, head of product and strategies at MariaDB. “The front end is fully SQL-compliant and the back end is ACID [atomicity, consistency, isolation, durability]-compliant with the same interface.” Vyas said the OLAP engine has been in development for “eight or nine years.”

The ability to support column stores should enhance MariaDB’s analytical appeal. Transactional databases typically store data in rows because most queries are looking for multiple fields of data about the same record. However, row processing is inefficient for analytical applications, which are more likely to seek the same field across multiple records for comparison purposes.

The process of converting from row to column storage usually involves an extract/transform/load (ETL) process, which is both cumbersome and time-consuming. While ETL is still required when multiple data sources are involved, users who want to work exclusively with transaction data in the relational store will gain both performance and convenience from the automated transformation.

“Customers hate the complexity of having to go to other vendors to purchase appliances to perform analytics,” said Lilia Shirman, MariaDB’s interim chief marketing officer.

The automated column store transformation feature will be exclusive to MariaDB, but the company is shipping a set of Open Database Connectivity (ODBC) and Java Database Connectivity connectors that customers and third-party vendors can use to integrate with their reporting engines. Developers are also free to write their own open source extensions. ANSI SQL compliance makes accessing the data store straightforward for most business intelligence tools.

The new streaming capabilities in MariaDB MaxScale are said to simplify real-time data propagation to external data lakes or data warehouses.  MaxScale and MariaDB can be configured to replicate without affecting performance, while including all necessary metadata so that the source data can be ready by any program without additional overhead.

MariaDB ColumnStore will be sold as a separately priced product, but pricing has not been set. The MariaDB engine is currently priced on a per-node basis at prices of between $6,500 and $7,500 per node, Shirman said.

Photo by Beatnik Photos via Flickr CC