SAN FRANCISCO — Oracle founder and CTO Larry Ellison this week detailed the company’s autonomous Oracle Database 18c for the cloud, which Ellison said will rely on advanced machine learning techniques to greatly reduce database administration tasks, such as tuning and patching.
At the heart of this Oracle cloud database is extensive use of machine learning, which Ellison called “the first branch of artificial intelligence that really works.”
This application of machine learning, which employs neural networks and other modeling algorithms to sift large amounts of log data and detects recurring patterns of database activity, is also part of a cybersecurity product for automatically patching databases that Ellison pledged to discuss further at the event.
For Oracle Database 18c, machine learning allows it to “patch itself while running, all without any downtime whatsoever,” according to Ellison, who spoke at Oracle OpenWorld 2017. He also said operations improvements allow his company to offer an Oracle Database 18c SLA that guarantees 99.995% reliability and availability, while reducing planned and unplanned downtime to less than 30 minutes per year.
Automation of database administration
While, the system, which Ellison dubbed as “self-driving” and “the world’s first autonomous database,” may be unique by some measures, it is also part of a long-standing trend that is well under way.
Automation of database cluster deployment on cloud has become increasingly common, and wider automation can be anticipated, according to Tony Baer, an analyst at Ovum.
“You can see how cloud databases are doing automation — with database sharding as a major example,” Baer said. Meanwhile, query performance and other database activities are also being affected by advanced machine learning technology, he said.
Baer noted that “Oracle has all kinds of database activity logs. That is big data that acts as a corpus for machine learning that can figure out what is a normal pattern, and highlight queries that are going to cause trouble.”
Advanced machine learning adds another element to the mix, but the latest Oracle moves are best viewed as part of an evolution in process automation, according to Vinod Bhutani, database services manager at DBAMart Database Services in Broomfield, Colo.
“There is a whole lot of automation for the database already. For example, there are such tools as Oracle SQL Tuning Advisor and Segment Advisor,” Bhutani said in an interview at Oracle OpenWorld.
“In my view, the database is 60% to 70% automated already,” he said, adding that the amount of automation employed is often based on the database administrators’ comfort levels with such automation’s effectiveness.
Bhutani said he would be looking for additional details, particularly on Oracle’s cybersecurity offerings, to see how much further Oracle takes database automation.
Whither the DBA?
In his Oracle OpenWorld keynote, Ellison admitted the move to greater automation for the Oracle cloud database could be seen as a threat to DBA job security. But he was basically sanguine on the prospects.
“Yes, you are automating the ways of database professionals, but they already have more work than they can possibly ever get to,” he said.
Greater database automation will free up DBAs from routine patching and repetitive tuning, he said, enabling them to focus more on schema design, analytics — including advanced machine learning styles of analytics — and securing data.
Noel Yuhanna, an analyst at Forrester, on hand at Oracle OpenWorld, agreed. “The DBA job is being changed toward more data-driven initiatives, with more emphasis on security and governance — and architecting the future of the data,” he said.
“The DBA will focus more on business value, as opposed to technology,” he said.
Meanwhile, analyst Baer also pointed to an increasingly important role for the DBA. “There is definitely a future for the DBA. There is just no question about it,” he said. “You can’t automate everything.”
Hearing Redshift steps
Ellison said the Oracle Database 18c, running on Exadata infrastructure on the Oracle Cloud or Cloud at Customer , will become generally available in December for data warehousing only, with a transactional version appearing in June of 2018.
This “data warehouse-first approach” emphasizes Oracle’s intention to compete more fully with Amazon, its cloud and its Redshift cloud data warehouse. At Oracle OpenWorld, Ellison repeatedly cited Redshift as a competitor, claiming superior uptime and better relative pricing for Oracle.
“We guarantee our bill will be less than half of what Amazon Redshift will be,” he said. “We will write that in your contract.”
The company further moved to sweeten its cloud pricing deal recently, introducing a “bring your own license” policy for existing customers moving databases, middleware and more to the Oracle cloud platform.
With the “18c” designation, Oracle takes on a model-year style naming format for its database, not unlike that of Microsoft SQL Server. Aligning database naming with calendar years is in some part a bow to the growing use of yearly, subscription-based pricing models for databases on the cloud.