This content is part of the Conference Coverage: OpenWorld 2017 in review: Oracle cloud offerings at center stage

Machine learning in Oracle's autonomous database not so magical

Oracle is employing a new autonomous database as part of its drive to the cloud. The technology relies on machine learning, but the actual effect is more evolutionary than revolutionary.

Cloud computing has become a very powerful force in the design of databases. In going to the cloud, users have had to take a brand new look at relational databases, and from a variety of angles.

Vendors have had to rethink those databases, too. Elastic scalability, automatic failover and on-all-the time operations are a part of that. Among the vegetables in the stew is a self-driving autonomous database.

This can lead to conversations that are, well, cloudy. A case in point comes from this month's Oracle OpenWorld 2017 in San Francisco.

At the event, Oracle co-founder and CTO Larry Ellison made the case for Oracle's commitment to success in the cloud. He said Oracle was employing new machine learning technology to better automate the maintenance and tuning of its new Oracle Database 18c software, which initially will be available as part of a cloud service. Also in the offing are cloud security products that will use advanced machine learning for anomaly detection.

Larry Ellison, CTO, OracleLarry Ellison

Certainly, machine learning, after some fitful starts, does seem to be ready to pay off in a variety of applications across the computer industry. But as Ellison rolled out the new autonomous database in his opening keynote, there was more than a bit of a hard sell going on.

How big is machine learning? To hear Ellison tell it, machine learning in cloud databases will be very big indeed. "I don't use the words revolutionary new technology every year here at Oracle OpenWorld because there aren't that many revolutionary new technologies," he told the audience. "But this one is."

Smoke, mirrors and cloud

If Oracle was somewhat behind others in pushing machine learning, Ellison seemed ready to make up for lost ground. Still, there is some smoke and mirrors in all this. That is because at least some of what Oracle is now set to provide is really just par for the course for a modern autonomous database, especially one in the cloud.

If Oracle was somewhat behind others in pushing machine learning, Ellison seemed ready to make up for lost ground.

Yes, machine learning is employed in the automation of upgrading, patching and running the new Oracle Autonomous Database Cloud service. However, it is less the revolutionary breakthrough Ellison portrays than it is a step in an evolutionary development that has been going on for some time -- albeit one that has gained momentum as DevOps and cloud computing take greater hold.

Among viewers that see a fair measure of hyperbole here is independent analyst Curt Monash. The competitive distinctions Oracle is trying to draw as it describes its new database are overly exaggerated, according to Monash. "Zero-tuning deployment has been an option for many database management systems, including Oracle's, for at least two decades," he said.

There are caveats, he added. For example, the performance that automated tunings obtain can't be mediocre. "It has to be very good," Monash said. "And now that is also the case with cloud integrations and cloud deployments. But Oracle's proposals don't sound revolutionary in either area."

DBaaS reboot?

It is not unfair to say that Oracle is a bit late to the database as a service (DBaaS) party, according to Warner Chaves, a Microsoft data platform principal consultant at IT services provider Pythian.

"Ellison made it sound in the announcement like they are breaking new ground here, but the truth is that some of their competitors already have years ahead of them in this area," Chaves said.

One cloud database competitor that Chaves points to is Microsoft Azure. "Azure SQL Database, for example, has always been patched under the covers without the clients doing anything -- not only for security, but also for enabling new features," he said.

Chaves also sees Oracle's load-and-go efforts being influenced by Google's BigQuery DBaaS, which has been commercially available for almost six years.

He does credit Oracle, however, with significant steps forward in its service-level agreement (SLA) policies for cloud databases. The company pledged a 99.995% SLA guarantee with the new autonomous database, pegging machine learning as one of the sources of greater availability. "Twenty minutes savings per year is no small feat," Chaves said.

"These are all awesome things that will heat up this space even more, and will drive innovation harder from all the big vendors," he added. "But let's not pretend Oracle is the first one bringing them to the table."

In a way, Ellison and Oracle have engaged in this before. Famously, cloud was vaporware in Ellison's estimation -- until he came down with cloud fever himself. As Gartner analyst Adam Ronthal remarked in a recent interview, "Oracle didn't like cloud until they invented it."

The company and its leader have a unique mix of engineering skills and showmanship, it is true. But conflating evolutionary advances with revolutionary advances doesn't help bring clarity to machine learning in cloud operations, or to the important decisions data managers have to make as they look to significant cloud migrations

Next Steps

Learn more about Oracle's and its competitor's database cloud services

See how web services paved the way for data in the cloud at Amazon

Take a look at the substantial and the specious at Oracle OpenWorld 2014 

Dig Deeper on Oracle cloud computing infrastructure