Speaking today at the opening keynote for Oracle Open World 2017, founder, executive chairman and chief technology officer Larry Ellison outlined Oracle's next major software release.
"The biggest threat by far in cyber security is data theft," Ellison said. "Preventing data theft is all about securing your data wherever you choose to store your data. We're going to make the case the safest place for you to store your data is an Oracle database, particularly an autonomous database," he said, setting the scene for his talk.
An autonomous database, Ellison explained, is highly automated. It will do everything it possibly can to avoid human intervention.
"Oracle has new cyber security technology to automate threat detection and then immediately remediate the problem. This new database technology means an automated database can instantly patch itself while running. There's no delay waiting for a human process, or a downtime window."
This is the autonomous database, and it's Oracle's vision for Oracle 18C which will be released for different database workloads from December 2017 through 2018.
Oracle Data Warehouse edition will be available in December 2017, Oracle OLTP database cloud service in June 2018, and Oracle NoSQL database cloud services by the end of 2018. Each edition will have autonomous functionality by default, and will all be available on-premises, in Oracle's cloud, and on public cloud services, to meet customer requirements.
The big thing about an autonomous database, Ellison explained, is the automation of cyber security and the automation of database operations working together. "The database has to be able to patch itself immediately while running when instructed by the cyber security system – no delay, it has to happen immediately, and has to happen without human intervention."
The technology driving autonomy is "as revolutionary as the Internet", Ellison said. "It's called machine learning, a branch of Artificial Intelligence, or AI.
"For years and years, AI did not live up to its promise. There is a new type of AI, machine learning. The applications are revolutionary. It powers self-driving cars, facial recognition better than humans can do, and new applications like an autonomous database and automated cyber security."
Machine learning is when computers learn from patterns in data and make predictions. It relies on large amounts of accurate training data. Higher volumes of accurate data increases learning, and increased learning means more accurate solutions to problems.
An example of machine learning is anomaly detection, to separate normal from abnormal patterns in data, like detecting cancer from normal cells or detecting the chief financial officer is logging in from a computer in the Ukraine and that is out of the ordinary.
In Oracle's case, the new autonomous database will be driven by machine learning absorbed from vast amounts of event logs including infrastructure logs — network, server, storage, VM and OS, platform logs — database, Java, analytics, and application logs, ERP, CX, HCM and more.
This event log training data will enable new machine learning applications across security, like detecting and correcting anomalous events be they logins or unusual SQL queries, and across database operations, such as classifying normal query patterns and automatically tuning the database.
On this point, Ellison says the Oracle 18C autonomous database will continuously tune the database based on information it gathers. "This is a big deal," he said, "Nobody else does this.
"Oracle 18C is the world's first 100% self-driving autonomous database," he said. "The world's first and only.
"This is the most important thing we have done in a long, long time."
Ellison continued to state the autonomous database will have total automation, based on machine learning. It would require no human labour to manage the database whatsoever. The database will automatically provision, upgrade, patch and tune itself while running; it will perform automated real-time security patching with no downtime window required and will have no human error.
The Oracle cloud SLA will guarantee 99.995% reliability and availability. "We will minimise costly planned plus unplanned downtime to less than 30 minutes per year," he said.
In one of many jabs at Amazon's AWS hosting services, Ellison said: "We guarantee your Amazon bill is cut in half, and the lower labour costs are even bigger savings. We will write that in your contract."
He said Oracle 18C would perform fully automated database provisioning and management, even for mission-critical scale-out clusters with data centre disaster protection. The user must define policies but the system automatically manages itself thereafter. It will perform automatic provisioning, back-up, upgrades, patching and tuning while running, and without any human administration needed, further cutting errors and malicious behaviour.
While this autonomy is the headline feature in Oracle 18C, Ellison noted it also includes a range of OLTP and analytics improvements, including 4x faster in-memory OLTP, 5x faster RAC for high-contention OLTP, 2x faster streaming ingress for IoT workloads, NVRAM ready row and column store, 2x faster in-memory column store, 100x faster in-memory analytics for external data, 100x faster approximate query processing, and more features.
Poking fun at Amazon's AWS hosting, Ellison said, "You get all this, but the shock is you must be willing to pay a lot less."
He went on to demonstrate the greater performance of running Oracle in Oracle's cloud vs. on Amazon's AWS cloud, and an even greater performance disparity compared to Amazon's RedShift. This performance improvement translates to real cost savings, which Ellison says, is further compounded by the reduction of human labour to run the database.
Ellison said Oracle cloud's reliability is 100% more reliable than Amazon with no exceptions in the fine-print, noting Amazon's contracted uptime excludes planned downtime, downtime for adding compute and storage capacity, downtime for Amazon planned maintenance, downtime for database upgrades and patching, downtime for regional outages, and downtime for software bugs.
"They may as well have excluded a few more things and just their uptime guarantee is 100%," he joked. By contrast, he stated, Oracle's uptime guarantee is 99.995% including mission-critical workloads, and without exclusions, and is guaranteed to be at least half Amazon's costs for a similar workload.
The minimum configuration for Oracle cloud is one OCPU, or Oracle CPU, roughly defined as one physical core of an Intel Xeon CPU with hyper-threading enabled, along with 1 terabyte of storage, for US$300 per month.
For database administrators, the automation features of Oracle 18C will allow them to turn their mind to innovation. "Database administrators don't lack things to do," Ellison said. Instead, with Oracle 18C administrators can spend less time on infrastructure, patching, upgrades, ensuring availability and tuning, and more time on database design, data analytics, data policies and securing data.
"Oracle 18C is the only self-administering database on the planet," Ellison said. "It's fully automated and fault-tolerant, with real-time security patching, constantly tunes itself to get the best performance, and it's fully automated and will go out and get additional compute resources the moment they're needed but not a moment sooner, so you only pay for what you use."
The writer is attending Oracle Open World 2017 as a guest of the company.