Oracle's Souder Talks Distributed Databases

After all, Oracle sure seems wants everything in the world to be in one huge (Oracle) database. At least that's what Microsoft and IBM say. That's even what Oracle chairman Larry Ellison also often says when he speaks about the company he founded.

But don't get Souder started on that subject, which he says competitors have blown out of proportion. "What we say is if you can consolidate information, you will probably be able to save money," he said in a recent interview. "There are many reasons you may not be able to consolidate--you may have legacy systems on a proprietary platform...I don't think there's a single large corporation today that runs on just one database."

IBM's federated data worldview, where data continues to reside wherever it already is and is pumped into a database when processing is needed, is therefore not unique, he maintains. IBM is pitching its upcoming DB2-based Information Integrator for that very purpose.

Souder insists that all database vendors support data synchronization technology that accomplishes that basic data consolidation. He also maintains, unsurprisingly, that Oracle is much better at it than the rest. Where IBM pitches a separate Data Joiner or Information Integrator, to accomplish this task, Oracle's database does all that synchronization work integrally, he said. "With Oracle you integrate the information and with the other guys you have to integrate their product and then integrate the information," he said.

He also insists that with Oracle's "location transparency" notion, available since Oracle 5.1, data can move around from system to system without need for changing the application.

Oracle, the market leader in distributed databases, is feeling increased competition from IBM's DB2 and Microsoft's SQL Server. Each company's claims and counterclaims about features and capabilities have reached a fever pitch in the past 18 months and it's extremely difficult to sort them out, said analysts.

Oracle has been hurt in the past by the tendency of its chairman to make derogatory claims about competitors. Two years ago, his contention that IBM's "shared nothing" architecture "does nothing," irked many in the analyst community who maintained that there are some tasks where shared-nothing architecture is extremely effective. A shared-nothing system divvies up data into sectors that are only accessible from pre-determined nodes. So if one node fails, that data becomes unreachable and the application fails.

"Look, Oracle's database is really very good. He should stick to what it does well and stop making claims about rival technology that are not necessarily true," said one database analyst who requested anonymity.

For more than a year, the database companies have claimed leadership in XML support, for example. In fact, all of the databases handle XML data and will add more native capabilities going forward.

According to Sounder, they all also will support the Xquery language for accessing XML data when it becomes available. That's not really grounds for competition. "Creating a new API on an existing mechanism is not truly a change in the landscape, it's just a new API," he noted.

"When the standards body writes a new data access [method] we will all implement it...it's not a fundamental change," he said.

What constitutes real innovation, he said, is the melding and merging of synchronous and asynchronous infrastructures, something he claims Oracle has long accomplished that in its "Streams" technology, embedded in the database.

What will be revolutionary in the future will be a way to dynamically provision data inside the enterprise--to come up with some way to balance the workload when one machine is idle and another is overloaded. All companies face such imbalances. In a manufacturing facility it could be an MRP run; in retail, it could be the quarterly close, when some machines are choked with work and others are underutilized.

"Companies are really interested in how to eliminate this imbalance....a way to flexibly allocate computing and data resources to match priorities they have [going on] right now and understanding how those things will change," he said.

Oracle's work in grid computing could play well here. Last November, Oracle announced its foray into the grid world, and launched an open source grid toolkit with Globus. (See story.)

Grid computing is the notion of dynamically breaking large computational workloads out of the data center and distributing them to myriad machines as required, typically over the Internet. IBM's computing on-demand initiative would rely on a grid model and would require users to pay based on computing capacity used.

The database world is certainly not standing still. This summer, Oracle is expected to announce and release its next generation database, Oracle 10i, observers said. Microsoft's next-generation SQL Server, code-named Yukon, is in limited beta now, and has been promised for general beta in the first half of the year. Microsoft officials last year said to expect delivery by the end of 2003, but are no longer saying that. IBM says its Information Integrator is in beta now and the first phase will roll out this year. A version with Xquery support is due next year.