Oracle Corp. today embellished the HeatWave database query accelerator it introduced last year with machine learning-based automation features that it says further improve performance.

MySQL Autopilot, which is being introduced today, includes automated provisioning, parallel loading, data placement, query plan improvement, scheduling and error recovery. Oracle said the automated features make the query optimizer increasingly intelligent as more queries are executed, thus continually improving performance over time.

Oracle said it’s targeting rival analytical engines, including Amazon Web Services Inc.’s Redshift and Aurora, Google LLC’s BigQuery and Snowflake Inc.’s namesake cloud data store. The database giant says the technology makes queries run between 10 and 1,000 times faster than rival databases at less than half the cost.

In line with that, the company rolled out benchmark results that it claims shows that MySQL HeatWave is 1,400 times faster than Amazon Aurora on some queries based on a test suite derived from the TPC-H benchmark.

60 patents

“As a cloud provider we believe there are a lot more opportunities for automation,” said Nipun Agarwal, vice president of research and advanced development at Oracle. “We have more info about the hardware and software and can do a better job.” The company said the latest enhancements are the culmination of nearly a decade of engineering that resulted in more than 60 patents.

For example, Agarwal said, “when a user wants to know what size to provision for a workload most cloud services expect the users to furnish that information.” With Oracle’s automated features, he added, “the system learns what you need. You only provide the tables or the schema and it calculates the rest.”

Oracle today also introduced MySQL Scale-out Data Management, which it said improves the performance of reloading data into HeatWave up to 100-fold. The maximum HeatWave cluster size has been increased to 64 nodes from 24 nodes and total processing capacity has been upped to 32 terabytes of data from 12 terabytes.

Oracle picked up MySQL as part of its $7.5 billion acquisition of Sun Microsystems Inc. in 2009. Although the project is open source, Oracle has a significant influence on which enhancements are included in the code as well as how quickly changes are made.

Many users of MySQL, which is part of the Lamp stack (from Linux, Apache, MySQL, PHP) that powers the majority of web servers, feared that that database management system would wither on the vine under Oracle’s stewardship. But many now admit Oracle has invested aggressively in the technology. With the introduction of a managed service, Oracle has an opportunity to monetize the engine, which is the second most-used relational engine on the market behind only Oracle’s namesake product, according to DB-Engines.

Faster loading

In addition to automating aspects of provisioning, MySQL Autopilot also speeds data loading by looking at running queries and partitioning to achieve the best performance. As data is loaded into a HeatWave cluster “we make a copy in-memory and keep it in the object store,” Agarwal said. “When we need to reload again for a restart, recovery or upgrade, the system doesn’t need to go to the MySQL database; it can read it from the object store.” That improves the load time for a 10TB database from about 7.5 hours to four minutes, he said.

The combination of analytical and transactional processing in the same engine continues a trend that has been ongoing for several years as vendors seek to make decision-support features available in their core engines without requiring extracts into a data warehouse. SAP SE’s S/4 HANA and MariaDB Corp.’s MariaDB ColumnStore are examples of other products that incorporate both types of processing.

Combining analytics and transactions in the same engine also enhances security by reducing attack surfaces, Oracle said. HeatWave communication is further encrypted using keys and certificates that are managed by an identity and access management service in the Oracle Cloud.

Oracle doesn’t see the managed service as competing with its flagship database, since the use cases are quite different, said Steve Zivanic, global vice president of database and autonomous services product marketing. MySQL is more common in large web serving environments whereas Oracle is typically used for transactions. Because MySQL is open source, he said, “our sellers didn’t previously focus on MySQL opportunities. Now they will.”

“There are greenfield opportunities like cloud native companies that have never had a native on-premises environment and many are running some form of MySQL,” he added. “It’s a net incremental business opportunity. Oracle cloud sellers have never had a solution like this.”

The company said all benchmarks it is using in the rollout were conducted independently and that the scripts are available on GitHub for users to test for themselves. The service is available today across all 30 Oracle Cloud regions. MySQL Database for HeatWave-Standard-E3 is priced at less than 34 cents per node per hour.

Image: Pixabay