Oracle CEO Larry Ellison is taking the fight to IBM, Microsoft and SAP in the burgeoning in-memory database market with a new option the company says can deliver dramatic performance boosts without requiring changes to applications.
Ellison first unveiled Oracle Database In-Memory at September’s OpenWorld conference. During a webcast Tuesday, Ellison is expected to announce the product will be generally available within 60 days.
That’s a bit sooner than some industry observers expected, and Oracle’s aggressive timeline speaks to the competitive pressures it is facing from rival in-memory database vendors. SAP, for one, has ported its Business Suite software to Hana, an in-memory platform it launched a few years ago, with the express goal of convincing Suite customers now using Oracle’s database to make the switch.
In-memory computing pushes information into RAM instead of reading it off of disk, providing significant performance boosts. Oracle has already offered the TimesTen in-memory database in both stand-alone form and as a caching option for its flagship database, but that technology has tended to serve more specialized purposes.
With the new in-memory option Oracle has an advantage over newer in-memory platforms such as Hana, said Tim Shetler, vice president of product management. SAP is “trying really hard to finish the other parts of a database management system,” he said. “It’s just a matter of maturity.”
In contrast, Oracle’s in-memory option can take advantage of the company’s well-established capabilities for database scalability, high-availability and security, Shetler said.
The in-memory option is only available with the enterprise edition of Oracle’s database, as is typically the case, Shetler said. Pricing won’t be released until the general availability launch but licensing will follow standard Oracle guidelines, he said.
Ellison is expected to show live demonstrations of the in-memory option running on multiple hardware systems, from a two-socket server to a full-rack Oracle Exadata machine and finally, a SPARC M6-32 server.
The demonstrations will show the software processing up to a trillion pieces of data, pulled from six months of Wikipedia queries, at high speed, according to Shetler.
Oracle took a best-of-both-worlds approach to its in-memory option. The add-on creates an in-memory column store, which dramatically speeds up analytic queries, while preserving the database’s existing relational row store for OLTP (online transaction processing) workloads. The column store does away with the overhead required to maintain row-based analytic indexes, improving OLTP performance.
Since the OpenWorld announcement, Oracle has added additional features to the in-memory option, the key one being support for its RAC (Real Application Clusters) technology, which provides scalability and fault tolerance.
Despite the option’s benefits for analytic processing, there’s still good reason for customers to buy Exalytics, Oracle’s dedicated analytics appliance, according to Shetler.
Exalytics should be seen as a dedicated platform for running Oracle’s BI (business intelligence) stack and Endeca data-discovery tool, with data being pulled into the system from multiple locations, he said. The in-memory option, meanwhile, helps speed up multiple applications that share the same database server, he added.
One observer gave the new option a measured thumbs up.
“The two standard concerns about Oracle are cost—product and administration alike—and scaling difficulties,” said analyst Curt Monash of Monash Research. “Both apply to this option in spades. But otherwise, it sounds great, for all the usual RAM performance reasons.”
Ellison, who is well-known for the barbs he throws at competitors, may not have nearly as kind words for rival products during Tuesday’s webcast.