Oracle adds autoML to its MySQL HeatWave service

As with analytics, so with ML – we're keeping everything in the box, says Big Red

Oracle has added autoML, real-time elasticity, and price-performance trade-offs to Heatwave, the analytics engine it piggybacks on MySQL transactional database.

Nipun Agarwal, veep of HeatWave and advanced development at Oracle, told The Reg that the autoML features included in Heatwave would handle model training, inference, and explanation on behalf of the users, saving customers time and resource extracting data for these purposes.

"Customers need to extract data out of the database, and then move it on machine learning," he said. "It is the same problem which customers had with analytics in the past, which is when you move the data out, the data is no longer as secure and it takes effort.

"It introduces complexity to the application. And finally, you're running this machine learning on some other service and it's going to cost you, and this is what happens for instance, with [AWS] Redshift ML.

"What we are introducing with HeatwaveML is customers get to run the entire lifecycle of machine learning from training to inference to explanation, all inside MySQL Heatwave, without the data or the model leaving the MySQL database," Agarwal said.

Introduced in December 2020, Heatwave was designed as an analytics system that MySQL customers could use without having to export data from their transactional applications to a specialist analytics system such as Teradata, Snowflake, or AWS Redshift.

Oracle argued customers save effort through simplicity and cost by avoiding data egress. However, it is not available to all MySQL users.

The open-source database was first developed by MySQL AB, which was bought by Sun Microsystems in 2008. In 2010, Oracle bought Sun, and MySQL co-founder Michael Widenius forked the code to a new open-source database: MariaDB, which has its own DBaaS offering and its own analytics story.

Although MySQL itself is available on Google Cloud, AWS (RDS), and Microsoft Azure, Heatwave is not. Agarwal said it was available on "commodity cloud," but for Oracle that means its own infrastructure, OCI.

Philip Carnelley, AVP of European software research at analyst IDC, said the move and the whole strategy with Heatwave was to avoid data, and money, leaking to Oracle's competitors when customers are running applications in MySQL.

"If customers are saying, 'We've got all this transactional data we need to analyze it as well' then it makes perfect sense for Oracle to take that money as well, rather than it going to their competitors.

"If people are already using MySQL, then it's making it easy for them to spend more money with Oracle rather than spend it with someone else. And, if you look at the way that the industry is evolving, it is encouraging people to have data in one place and run different services on it," he said.

Carnelley pointed out that, for example, SAP runs analytics on its transactional database in S/4HANA applications.

Also new to MySQL Heatwave this week is something Oracle calls real-time elasticity. Agarwal said that although the system was highly scalable, customers were required to increase the size of the cluster manually. The process has been automated and avoids downtime, he said.

Oracle was also allowing customers to set their own trade-offs for price performance. That is agreeing to less than maximum performance for a given system in exchange for lower prices. ®