3 min read

Oracle Open World: Key Questions On Cloud Vision

As Oracle CEO Larry Ellison lays out plans for infrastructure-as-a-service, Oracle-run Private Cloud services, 12C database, and Exadata X3, the questions start with scale.
"Pluggable databases are vastly more efficient; we can use one-sixth as much hardware [capacity] and run five times the number of databases," Ellison said. "With five times more scalability, that's important to conventional database users and SaaS companies."

12c will be generally available next year, Ellison said, but he didn't spell out the pricing or licensing terms. A key question is how will the 12c architecture change the Oracle database licensing scheme? How many private databases will a container hold? Will per-CPU charges go out the window in favor of fees based on the number of database containers and pluggable databases? Or both models apply, adding yet another layer of complexity to the database licensing scheme?

Exadata X3, Oracle's fourth major announcement, was touted by Ellison as the foundation of the Oracle cloud. "If you thought the old Exadatas were fast, you ain't seen nothing yet," he said.

A single X3 rack packs 4 terabytes (TB) of DRAM, 22 TB of flash cache, and, with 10X compression, can store a total of 220 TB of customer data. With all the DRAM and flash cache available, Ellison said it would be possible to keep all active data in memory so you "virtually never use disk drives."

Informing the audience that "disk drives are becoming passe," Ellison echoed SAP co-CEO Jim Hagemann-Snabe, who made that observation at that vendor's SAPPHIRE event in May. But Ellison's only reference to SAP was an unfavorable comparison of SAP Hana, which Ellison said offered only half a Terabyte of memory, to Exadata X3, with 26 TB of memory. (It's an apples-to-pears comparison, however, as Hana runs entirely on blade server DRAM and is scalable far beyond 500 gigabytes.)

In an upgrade to Exadata's architecture, the appliance's flash cache can be used for database writes, not just reads as in early appliances. This will improve write performance by a factor of 10 to 20, according to Ellison, and this software-supported upgrade will also be available to owners of existing Sun-hardware-based Exadata appliances, he said.

Comparing Exadata X3 to an EMC VMax disk array, Ellison reported that Exadata had double the I/O capacity of the EMC hardware with a single rack, yet Oracle could keep increasing capacity by adding additional racks whereas the VMax was "maxed out."

Exadata X3 was also compared to an "equivalent" IBM Power server deployment, with the Exadata reportedly delivering higher performance even though the cost was $650,000 versus $5.6 million for the IBM configuration.

As we've seen in the past, and as it has been admonished for in its advertising, Oracle tends to twist the facts; the questions on the Exadata X3 are all in the details. Exadata X3 is said to be the same price as the X2 in terms of hardware. With the pricing questions surrounding the 12c database possibly adding to the complexity, it will be some time before the basics of Exadata X3 cost-for-performance specs are known, let alone comparatively understood.

Larry Ellison's keynote was prefaced with the following safe harbor announcement: "This presentation should not be relied upon in making purchasing decisions." Truer words were never spoken.

In-memory analytics offers sub-second response times and hundreds of thousands of transactions per second. Now falling costs put it in reach of more enterprises. Also in the Analytics Speed Demon special issue of InformationWeek: Louisiana State University hopes to align business and IT more closely through a master's program focused on analytics. (Free registration required.)

Editor's Choice
Sara Peters, Editor-in-Chief, InformationWeek / Network Computing
Jessica Davis, Senior Editor
Richard Pallardy, Freelance Writer
Carrie Pallardy, Contributing Reporter
John Edwards, Technology Journalist & Author
Carlo Massimo, Contributing Writer
Salvatore Salamone, Managing Editor, Network Computing