Ellison Unveils Oracle In-Memory Analytics Server


Oracle Sunday unveiled a new business analytics server based on in-memory technology, a system CEO Larry Ellison described as having the ability to "deliver analysis at the speed of thought."

The product debut came during Ellison's opening keynote at Oracle OpenWorld, the company's massive annual customer and partner conference that has some 45,000 attendees this year.

Ellison devoted much of his hour-plus keynote to touting the capabilities of Oracle's "engineered systems" – Oracle's term for complete hardware-software systems. That lineup includes the Exadata database server, Exalogic middleware server, and the SPARC SuperCluster general purpose servers the company just introduced last week.

The new Exalytics Intelligence Machine incorporates the TimesTen in-memory database Oracle acquired in 2005. In-memory databases, which reside in a systems DRAM memory rather than on disks, can process data in near-real time. "There's nothing faster than this thing," Ellison said.

The Exalytics server also includes the Oracle Business Intelligence Extended Edition software and Oracle's Essbase OLAP (online analytical processing) multi-dimensional database. All that will allow the system to work with relational, mutli-dimensional and unstructured data.

The new product is competing with other in-memory computing systems including the High-Performance Analytics Appliance (HANA) system that arch-rival SAP has been selling since December 2010.

Recently a new generation of database software generally known as "No SQL" databases has been generating market attention, both because of their ability to handle unstructured data such as text and images, and because they can process huge volumes of data. Those new database systems include the open-source Cassandra database from the Apache Foundation and the MongoDB database.

The Exalytics server can be paired with Exadata to serve it with data for analysis, Ellison said. It also includes a heuristic adaptive in-memory cache that continuously decides what data needs to be stored in memory based on the analytic workloads.

The new server runs on Intel 10-core Xeon processors and has 1 TB of memory.