IBM has released a beta update to the DB2 database that will add the ability to store unstructured XML data separately from conventional relational data.

The upgrade, codenamed Viper, will be released in the middle of next year, according to database marketing director Bernie Spang. Spang said Viper will be able to store multi-media files, Excel spreadsheets and Word documents in an XML repository that will work in parallel with IBM's relational data repository under a single DB2 engine.

Relational databases typically handle XML data by storing the entire file as an object that isn't relationally indexed, or by "shredding" the file so the unstructured information fits into multiple relational data cells.

"Offering native XML functionality is very essential to delivering improved performance for data access," said Forrester analyst Noel Yuhanna. Oracle had "a head start on XML for many years," he added. "I expect this to become a game of catch-up and leapfrog among the big vendors."

Oracle officials said the company began offering XML storage options five years ago and in July enabled users of its 10g database to natively search XML files using the XQuery markup language. IBM's native XML storage feature "doesn't add any value", said Oracle XML technology manager Mark Drake.

Viper will also support XQuery for processing XML data, along with standard SQL. In addition, it will be the first DB2 release to support three different partitioning methods: range partitioning, multi-dimensional clustering and hashing. That support is aimed at helping IBM compete against Oracle, which has said that 10g offers six methods of partitioning data tables for faster access to information.

Despite estimates that the amount of unstructured or XML-formatted data at most companies is already larger and growing faster than structured relational data is, Yuhanna said he doesn't think that XML will replace SQL as the preferred data format.