Oracle Exalytics Aims for Business Intelligence at the Speed of Thought
Updated · Oct 03, 2011
Business analytics at the speed of thought? That’s the pitch that Oracle CEO Larry Ellision made at the Oracle OpenWorld conference with the announcement of the new Exalytics appliance.
Exalytics is an in-memory database appliance that leverages parallelism and DRAM to accelerate business intelligence (BI) queries.
“Everything runs faster if you keep it in DRAM, when you keep it in main memory,” Ellison said. “So we decided to put together a piece of hardware and a piece of software where the database is parallel and in–memory and the machine itself is all fully parallel.”
Inside the Exalytics box is 1 TB of DRAM for main memory. Ellison noted that Exalytics compresses data by as much as ten times. As such, after compression there might be 10 TB of data that has been compressed to 1 TB of physical memory.
“Hardware and software engineered to provide data analysis at the speed of thought,” Ellison said. “The analysis is instantaneous, as all the data you’re analyzing is in main memory.”
The Exalytics is powered by 40 CPU cores and can be connected to an Exadata database machine via a pair of 40 Gbps Infiniband connections. From a software perspective, Ellison explained that Exalytics leverages a new version of the TimesTen in-memory database that has been fully parallelized. He noted that TimesTen is a relational database which is being complemented with the multi-dimensional Essbase database that has been optimized for in-memory.
“The Exalytics machine not only handles relational data, not only handles multi-dimensional data, it also handles and analyzes unstructured data at the same speed,” Ellison said. “There is no response time, everything is pretty much instantaneous.”
Ellison stressed that Exalytics will run all existing Oracle BI applications without change, though he did note that Exalytics provides an interface re-design in order to facilitate the instantaneous response that the hardware delivers.
The ability to deliver the instant data analytics is dependent on having the right data in memory first. To that end, Exalytics has something called the heuristic adaptive in-memory cache.
“As different people ask different questions we migrate different data sets into memory,” Ellison said.
According to Ellison, running Exalytics on top of an existing Oracle database deliver analytics that is 18x faster. When running with an Exadata database machine, queries run 23x faster.
Oracle also unveiled the Oracle Big Data Appliance, an engineered system that includes an open source distribution of Apache Hadoop, Oracle NoSQL Database, Oracle Data Integrator Application Adapter for Hadoop, Oracle Loader for Hadoop, and an open source distribution of R.