Furthering its federated approach to data management, IBM Corp. is working to broaden an optional feature in DB2 7.1 Universal Database (UDB) that extends IBM's DataJoiner to Sybase Inc.'s database offerings and Microsoft Corp.'s SQL Server.
The extensions will emerge in a pending, unnamed version of DB2, which IBM plans to ship in approximately two months, said Jeff Jones, IBM's senior program manager of the data management group. "Our whole philosophy is wrapped around federation and integration -- not centralization," Jones said.
Although DataJoiner already provides links to Oracle's relational database, Big Blue and Oracle are competing not only technologically but also conceptually for the best way to manage data. Oracle is taking the other side in the debate.
"We don't believe in fragmenting data. You can call it federated, but the other way to say it is fragmented," said Jeremy Burton, senior vice president of marketing at Oracle.
IBM, however, maintains that enabling customers to manage a variety of databases from within DB2 is the best route because few companies have just one database vendor, Jones said.
A significant advantage to the federated approach is that users can sidestep the need to migrate data from a variety of sources, such as legacy and nonrelational systems, into a single repository. Whereas migrating small amounts of data is not problematic, moving a multiterabyte data warehouse is nothing short of Herculean.
Instead, IBM's approach extends the core database engine capabilities to sources outside the database, such as non-IBM databases. The forthcoming version of DB2 will extend to Sybase's relational database offerings and Microsoft's SQL Server, and Jones said that future iterations will support more data sources.
Oracle is not ignoring this need to access data outside the database engine. Companies that have content residing on the Internet, for instance, can use the database's query engine to access and index that content, Burton said.
Burton added that centralizing data makes it easier to manage.
But analysts said that IBM's federation provides the best of both worlds.
"IBM probably has the better approach because you can either use federation or you can bring all the data into the database to centrally manage it because it scales," said Peter Urban, an analyst at AMR Research in Boston.
The British Library, in London, which has a mix of databases as well as nonrelational and unstructured data is subscribing to the federated approach.
With more than 150 million items to catalog and archive, the library is tying together all the traditional material, including recently digitized medieval manuscripts, with the library's burgeoning collection of digital content -- and making it easily accessible via a single interface, said Helen Shelton, deputy director of collection management.
The first step of the federated approach is enabling visitors to search digitized archives using monitors in the library. "The vision ... is that everyone will be able to access the British Library's entire collection via the Web," Shelton said.
IBM fuses data sources
Users will have several concerns to weigh when IBM rolls out an upgraded DB2 that promises a single interface to major, non-IBM relational databases.
When dealing with huge amounts of data, users will not have to migrate data from existing non-IBM databases and legacy systems.
Queries against a federated system will run as if only a single data repository is involved.
DB2 users can take either the federated approach or pull data directly into the IBM database.
Depending on where data resides, querying can be slow.
A centralized approach is better for smaller data sums.
A federated approach may be tougher to manage than a centralized one.