31 July 2010
Saturday
Oracle to face off with IBM, Hitachi with Exadata 2
By Beth Pariseau | Sep 21, 2009
0 comments
Facebook
LinkedIn
Digg
Email
PrintOracle Corp. Monday night launched Exadata 2, its first joint product with Sun Microsystems Inc. since Oracle said it would buy Sun.
The product is specifically meant to support online transaction processing (OLTP), putting it squarely into competition with tier 1 storage vendors. Oracle CEO Larry Ellison provided plenty of rhetoric during the Monday launch webcast, calling out specific rivals in the high-end storage and server market, including IBM and Hitachi Data Systems (HDS). Meanwhile, previous hardware partner Hewlett-Packard (HP) Co. is out of the mix, according to an Oracle FAQ.
Oracle has not yet completed its $7.4 billion acquisition of Sun. European Union regulators are scrutinizing the deal, first revealed last April. But following a full-page ad in The Wall Street Journal last week quelling speculation that Oracle planned to discontinue or spin-off Sun's hardware products, the two moved forward Monday night with the joint announcement of a new database machine based on Oracle software and Sun hardware .
The Exadata 2 is the successor to the Exadata system launched with HP last year. Ellison said the Exadata 2 is optimized for the random I/O associated with OLTP, as opposed to Exadata 1's strong suit of sequential I/O for data warehousing.
Exadata 2 consists of an eight-server compute grid with 64 Intel Xeon processor 5500 compute cores and 400 GB of DRAM, a Sun-designed InfiniBand switch, and 14 Sun storage servers containing a total of 5 TB Flash and 336 TB hard disk capacity. The previous version was based on HP server hardware and a Voltaire InfiniBand switch. Ellison claimed the new Exadata can reach up to 1,000,000 random read/write IOPS.
Exadata 1 had some intelligence at the storage layer, most notably the ability to process a partial result of a query before feeding data to the compute layer. But Ellison put even greater emphasis on Exadata 2's storage features, including what he called "a highly intelligent memory hierarchy" between DRAM in the compute nodes and Flash modules in the server nodes, which he was careful to point out was "not just a dumb Flash disk."
This article originally appeared on SearchStorage.com
