IBM Lets Loose Viper Hybrid DatabaseIBM Lets Loose Viper Hybrid Database

IBM said the world's first major hybrid database system, DB2 9, also known as Viper, will ship July 28 and will be able to handle the major form of unstructured enterprise data, XML documents, as well as structured data in traditional relational tables. The Oracle database system and Micrsoft SQL Server also handle XML data but IBM says DB2 9 is different. It will be the industry's first major system to allow seamless function between relational and XML-heirarchical data handling. With Viper, developers should be able to produce applications that direct DB2 to store and retrieve data without indicating whether it is structured or unstructured, said Bernie Spang, IBM's director of data servers. "It's a significant time and cost savings for developers," he asserted.

The same capability will provide users with "uninhibited information access," Spang predicted. Just as structured data is easily organized and found through relational system indexing, XML data can be indexed in DB2 as well.

Relational systems, including IBM's, have claimed XML-handling for years. But they accomplished it by breaking XML forms and documents down into relational tables, then reconstructed the whole document when XML information was sought. By handling XML in what IBM calls its patented "pureXML" format, a DB2 user could form a query using XML's standard query language, XQuery, and go straight to a piece of information in a document without retrieving and rebuilding the whole document.

The process saves CPU cycles, while delivering information faster to end users, Spang promised. He cited anecdotal feedback from beta customer AutoZone, an online auto parts supplier, indicating an XML query would proceed 40% faster than before.

IBM has talked about Viper's capabilities before, particularly its XML handling. But on the eve of its delivery, Spang spoke of its hybrid nature as "the biggest database advance in twenty years."

Wayne Kernochan, analyst at Infostructure Associates LLC, a software consulting firm, said the difference between IBM and competitors' approaches is that "IBM makes a great effort" to define and recognize XML data based on its unique characteristics, while other systems store XML data "as an undifferentiated mass of objects."

Because DB2 can recognize the nature of the XML data it is dealing with, it can optimize transaction performance for different types of XML, Kernochan wrote in an email response to a query.

"Does this constitute a two-decades-worth big idea? maybe not," he continued. But the way users adopt it over time may indicate that it was a breakthrough for a major DBMS to build in efficient XML-handling characteristics, he wrote.

DB2 9 will also offer a new form of data compression, nicknamed Venom, which provides a row-based method of compressing data objects. The capability allows a finer grained analysis of data than was possible before to eliminate repetitive patterns, particularly in large tables, Spang said.

In its announcement, IBM pressed its case that the changes in DB2 9 differentiated it more than previously stated from its competitors. The Venom compression method "is in stark contrast" to other vendors' older, table-based compression capabilities, the announcement said. In previous descriptions, IBM stayed focused on how Viper would use less disk space.

In addition, IBM claimed improved database security in DB2 9. Viper will offer label based access control, which allows database administrators to define particular data as off limits to all but the highest data access authorizations. Customer data might be generally available to sales force representatives, for example, but customer Social Security numbers within the data table could be restricted to only a few fully authorized company officials, Spang explained.

DB2 9 is also designed as more of a self-governing, self-fixing database system. It can self-tune the amount of memory assigned to a database, saving a DBA the task of reconfiguring DB2 servers with the proper configuration parameters and buffer pools. Likewise, the system can automatically grow the size of the storage system available to a running database. The system collects statistics on its own operations to help provide a basis for formulating rules and governance policies by which it's run.