Oracle Corp. CEO Larry Ellison appears to be taking a lesson out of the playbook of his buddy, Apple Inc. CEO Steve Jobs.
Ellison tantalized Wall Street this summer with promises of a "major database innovation" to be announced this month, presumably during the company's annual OpenWorld conference in San Francisco, which starts Sunday.
"It is going to be a very big and important announcement for us, so we are not standing still in database," Ellison said.
During its first quarter earnings conference call today, Oracle President Charles Phillips said the company will deliver an update to 11g this quarter, and confirmed that a database-related announcement would be made at OpenWorld, but did not provide specifics.
At the same time, Oracle has been more secretive than normal, presumably to build interest in this upgrade, Release 2 (R2) of its year-old 11g database.
The number of companies invited to the beta program is smaller than in years past, sources said. Those who did participate uniformly declined to reveal details about R2's features via their blogs or to reporters.
Moreover, in at least four cases, information about 11g R2 that was posted to the Web by Oracle or outside bloggers has been taken down, though evidence in Google Inc.'s search cache often remains.
"I've never seen anything like this," said Paul Vallee, CEO of The Pythian Group, an Ottawa, Ontario-based database services provider. His company was not invited to participate in the 11g R2 beta, despite having participated in every other recent one, he said.
So, what new features will arrive in 11g R2?
An Oracle spokesman declined a request for comment. But based on information gleaned via Web-based research, analyst interviews and Oracle's own prior statements, Oracle 11g R2 should include improvements in the following, somewhat overlapping areas:
Oracle Database is probably the most feature-rich database around, rivaled only by IBM's DB2. Yet, it possesses what some analysts consider an Achilles' heel -- its lack of true grid capabilities.
Oracle Database, along with Microsoft Corp.'s SQL Server, Sybase Inc.'s Adaptive Server Enterprise and a few other relational databases, was built to have a "shared everything" architecture, said independent database analyst Curt Monash.
Typically in that design, one instance of an application is spread across multiple CPUs or servers, which share a common pool of memory and disk storage. The advantage is that users can "scale up" their applications quickly if demand arises, added Pythian's Vallee.
"You can take an application from Windows to an IBM z mainframe with literally no code changes," he said.
The downside is that Oracle's Real Application Clusters (RAC) do not easily allow multiple instances of the Oracle Database to be coordinated and run in parallel on hundreds or thousands of cheap PC servers. That sort of massive parallel computing (MPP), which Monash called a "shared nothing" architecture, has been the trend for almost a decade, especially among Internet companies with huge data centers, such as Google Inc.
Not everyone agrees that shared-nothing architectures perform better than shared-everything ones, such as Oracle's. Kevin Closson, a performance architect in Oracle's Systems Technology Group and former chief technology officer at storage vendor PolyServe Inc., vehemently defended the "shared everything" approach in his blog.
But the net result is that Oracle "is way behind in the 'scale-out' world," Vallee said. "MySQL is eating its lunch in terms of Internet-scaled deployments."
The irony is that Oracle added the "g" to its database name starting with Version 10, released in 2003, in order to imply the product's grid-worthiness.
Oracle is aware it is behind. At the launch of 11g in July 2007, Andy Mendelsohn, Oracle senior vice president for database server technologies, said, "We're doing a lot of work in grid technologies for the next release, which will make grid infrastructure even easier to adopt."
Oracle has already brought out new grid features through acquisitions, such as its purchase of Tangosol Inc. in 2007. It will have a number of sessions at OpenWorld on the Coherence product acquired from Tangosol.
The company will also hold a session entitled "Oracle Grid Computing 2.0: A Preview" on Monday.
Monash isn't convinced that Oracle will make the overhaul to its shared-everything architecture that he said is needed, but others who claimed to have been briefed were more positive.
"We the database people will be living in the new Grid 2.0 era," wrote H. Tonguc Yilmaz in his Oracle blog.
"The changes to RAC and the grid as a whole are very exciting," wrote "Grumpy DBA" blogger Jay Caviness in July in an entry that was deleted sometime after mid-August. Caviness, in an e-mail today, said Oracle asked him to take down the post, saying he was bound by a confidentiality agreement and could not comment until the official announcement of 11g R2 database.
Clustering and storage management
There's more concrete evidence, however, for improvements in the narrower area of how 11g R2 physically stores data.
According to a presentation last year by Oracle vice president for database product management, Mark Townsend, 11g R2 will include the ability for "plug and play clustering." That means the ability to quickly bring additional servers or hard disks online without needing to rewrite, reboot or load balance.
According to information posted by Oracle service provider Burleson Enterprises Inc., "there are some super-important changes to storage management in Oracle 11g release 2."
The new features in Oracle's Automatic Storage Management (ASM) include the ability to take snapshots of data for backup purposes, rebalance I/O, place data in faster parts of the disk and unify the storage management, according to Burleson.
The Web pages on ASM were deleted sometime after Sept. 4. Oracle did not contact Burleson to remove them, according to Robin Rademacher, operations manager at the consulting firm, nor did Burleson remove them because they were inaccurate.
"We knew they weren't supposed to be published yet, that's why we took them down," she said via phone today. "We built them as placeholders knowing that R2 would be coming at OpenWorld."
Oracle officials are also scheduled to discuss a "new database accelerator" during three OpenWorld sessions on Sept. 25, as well as in a separately titled technical overview session later that day.
The frequency "sort of implies that we're talking about an actual new product here rather than just a marketing exercise," wrote Mark Rittman, a U.K.-based Oracle data warehousing and business intelligence specialist at Rittman Mead Consulting, on his blog.
"Supporting ever-larger databases, with ever-increasing demands for getting 'answers' faster and faster requires a new way to approach the problem," the description of the technical overview session, boldly subtitled "Query Processing Revolutionized," reads in part. "This session looks at the intersection of database, data warehouse and storage solutions that will deliver on these requirements."
Rittman speculated that the accelerator has something to do with Oracle's Optimized Data Warehouse initiative and possibly also involves partitioning, OLAP, Automatic Storage Management (ASM) or "some new 11gR2 feature that [Oracle executives] Chuck Rozwat or Andy Mendelsohn will announce earlier in the week."
A person familiar with Oracle's plans confirmed Rittman's speculation that Oracle plans to announce important initiatives in data warehousing appliances, storage and query acceleration.
The sessions are scheduled on the last day of OpenWorld, a day after Ellison's keynote address, entitled "Extreme. Performance."
Lucas Mearian of Computerworld contributed to this story.