Ellison: Oracle's In-Memory Tech Is So Fast, Some Customers Think Their Database Is Broken


Oracle unveiled its long-awaited in-memory database technology Tuesday, marking its official entry to a red-hot market that's all about speed and the near-instantaneous retrieval of data.

In an event at Oracle's headquarters in Redwood Shores, Calif., CEO Larry Ellison said Oracle's new "in-memory column store" technology, which works in conjunction with the vendor's 12c database, enables database queries and analytics to run between 100 and 1,000 times faster than in the past.

With in-memory technology, Oracle 12c database allows each CPU core to scan 2.5 billion rows per second, Ellison said.

[Related: Sources: Hewlett-Packard In Talks To Acquire Hyper-Converged Infrastructure Startup SimpliVity]

Oracle is harnessing that extra horsepower for its own applications. A customer receivables management app from Oracle JD Edwards has seen an eye-popping performance boost from using in-memory technology.

The time it takes for the 12c database to process 10 million invoice lines has been shrunk from 244 minutes to 4 seconds, Ellison said.

Oracle's in-memory technology performs queries and analytics so quickly that some customers in beta testing mistakenly concluded that their database was somehow "broken," said Ellison.

The extra performance means Oracle customers will be able to do business much faster than in the past, according to Ellison. "Enterprises can think differently about how they operate because of these technological changes," he said. "They can get information instantaneously now."

Oracle is coming late to the in-memory database space, where products such as SAP HANA and HP Vertica are already well-established. Ellison said one reason Oracle took longer is because it wanted customers to be able to use its in-memory database without having to make changes to their existing apps.

"All your SQL runs unchanged. There is no loading and unloading of data -- it just works," Ellison said of the in-memory database technology.

Ellison said Oracle also wanted to solve challenges that existing in-memory databases haven't yet conquered. One example is online transaction processing.

While in-memory database technology in general is designed for faster execution of database queries and analytics, current products on the market don't extend this to online transaction processing, Ellison said.

Online transactions run faster when data is stored in row format, while analytics run better when stored in column format, Ellison said. Oracle's 12c database stores data in both row and columnar formats, and that allows online transaction processing to run twice as fast as previous Oracle databases, he said.

Ellison said another issue with existing in-memory databases is that they don't scale well. Oracle has tackled this challenge with a technology called Real Application Clusters, which stores data across multiple cluster nodes, allowing it to take advantage of the memory from many different machines, Ellison said.

NEXT: More In-Memory Database Features Oracle Is Touting As Game-Changers