Database Race Is on to Scale Up

In the database world, as in many other areas of IT, companies are searching for ways to do more with less.

In the database world, as in many other areas of IT, companies are searching for ways to do more with less. This challenge is being approached from different directions by the three top database vendors.

Oracle Corp. is focusing on clustering and adding nontraditional database services such as file serving to its database to lower the cost of ownership. IBM is beefing up XML (Extensible Markup Language), warehousing and other integration technologies to make it easier to use DB2 to connect different databases (even non-IBM databases) as well as working on self-management features. Microsoft Corp. continues to have the simplest database to administer, which also includes strong data analysis tools.

Clustering has become a key battleground in the database industry. "Clustering allowed us to really tackle the economy issue," said Oracle CEO Larry Ellison, speaking at the launch of Oracle9i earlier this month.

Oracles OPS (Oracle Parallel Server), which was renamed Real Application Clusters in Oracle9i, automatically provides a fault-tolerant database design, but until this month, it had a fundamental technical flaw that kept it from scaling well. This flaw was that OPS used its shared disk system instead of a network link to communicate among nodes. With Oracle9i, all communication now happens over a network, although data is still accessed using a shared disk.

In contrast, both IBM and Microsoft use a shared-nothing design, where a group of databases, each with its own disk and memory resources, exchanges information over fast network links, a process that is substantially faster than using a disk to communicate and can run well on clusters of 32 or more machines.

However, IBMs and Microsofts designs dont automatically include fault tolerance, although they certainly both support it. They also arent supported by complex packages such as SAP AGs R/3 (another strength of Oracle9i clusters), although many applications, including those developed in-house, do work well on IBMs shared-nothing clusters. (Microsofts clusters are not yet ready for widespread use.)

Providing more comprehensive and less expensive data analysis services will be a focus for database vendors in the next few years.

Microsoft has been a leader in OLAP (online analytical processing) and data mining with its database, as well as top-quality data transformation and loading tools.

DB2 also includes a built-in OLAP engine based on Hyperion Software Corp.s Hyperion Essbase and, new in Version 7.2, includes a data mining extension as well. (For eWeek Labs review of DB2 7.2, see Page 55).

Oracle has sold a separate OLAP server for some time, and, new with Oracle9i, is selling OLAP and data mining extensions that integrate directly with Oracle9i.

Oracle, IBM and Microsoft are all racing to add XML features to their databases to strangle the native XML databases market before it gains a lot of traction; Oracle9i, DB2 7.2 and Microsofts SQL Server 2000 all provide a variety of ways to import, store, search and export data in XML format, and all three are moving quickly to enable remote access to database services through the XML-based Simple Object Access Protocol.