Oracle is moving into in-memory analytics, taking on EMC, IBM and SAP with its own appliance that integrates hardware and software to process data and give answers to enterprises.
Oracle's Larry Ellison announced the company's move into in-memory analytics at Oracle OpenWorld in San Francisco. Photo credit: Jack Clark
The Exalytics appliance, revealed by Oracle chief Larry Ellison on Sunday, is designed to run business intelligence analytics at high speeds via a terabyte of DRAM for in-memory computing.
"[Exalytics is] hardware and software engineered together to deliver data analysis at the speed of thought," Ellison said in a keynote speech at Oracle OpenWorld in San Francisco. "Everything runs faster if you keep it in DRAM, if you keep it in main memory."
Although storage costs are notably volatile, DRAM costs around $10 (£6.40) a gigabyte, compared with flash at around $1 a gigabyte and hard disk at 4 cents. However, DRAM has the advantage of allowing data to be processed and the results sent to the consumer at speeds that are orders of magnitude faster than data stored in other technologies.
The Exalytics hardware is powered by four 10-core Intel Xeon E7 processors with connectivity through 40Gb InfiniBand, and 10Gb and one-gigabit Ethernet. Ellison said that after data compression, the terabyte of DRAM reserved for in-memory computing could actually represent five to 10 terabytes of data.
The machine can scan all data in DRAM data within five seconds, Ellison said.
Whatever you want to analyse, you can compress and analyse in-memory with the Exalytics machine. – Larry Ellison, Oracle
Pairing an Exalytics appliance with an Exadata or Exalogic box can boost performance when running Oracle applications, he added, though Exalytics will work with non-Oracle workloads as well.
"Whatever you want to analyse, you can compress and analyse in-memory with the Exalytics machine," he said.
Beyond the hardware, Exalytics makes use of an updated version of Oracle's TimesTen relational online transaction processing (OLTP) database and the EssBase database to process structured, semi-structured and unstructured data. The analytics algorithms used by the database-querying engines have been rendered parallel, Ellison said, so they can get higher performance out of the Exalytic's 40 cores.
TimesTen's user interface has been "redesigned for instant response", Ellison said. This means the software can respond to queries as they are being typed, similar to how Google offers suggested searches when a phrase has been partially typed.
The launch is part of Oracle's nascent 'Parallel Everything' architecture. With this, Oracle is building parallel-friendly hardware infrastructure to support its increasingly parallel software applications, such as Exalogic for its middleware and Exadata for its database software.
"Database performance is all about moving data, not doing arithmetic in a microprocessor," Ellison said.
The basic idea behind Parallel Everything is that Oracle integrates hardware and software, and designs both for a common goal. "You can't have a great parallel database computer unless you have great parallel storage servers to feed it," he explained.
Ellison also used his keynote to answer criticism of Oracle's hardware strategy in the wake of its purchase of Sun.
"Every night the sun sets; but the sun also rises," he said.
Exalytics will launch into a crowded market, with existing and upcoming products from major companies — IBM has its Netezza appliances, EMC its recently acquired Greenplum data computing appliance division and SAP its Hana scheme.
Ellison did not give prices or expected shipment dates for the Exalytics appliance.