Analysis: Microsoft raises the database stakes

SQL Server 7.0 is a critical release for Microsoft. Previous releases of the database, although widely used, have never gained a reputation for mission-critical performance and scalability.

Oracle, in recent months, claims to have retaken the lead in volume shipments of Windows NT databases. Other vendors ? notably IBM and Informix ? are making inroads into the high-end of the NT market. SQL Server 7.0 must deliver a total package if it is to compete successfully. While the overall quality of the release has yet to be decided through real-world experience, it would appear that it delivers.

SQL Server has always been a leader in out-of-box usability. Version 7.0 extends this with performance analysis and tuning wizards which reduce the need for expensive database administration skills. Visual Database Diagrams provide a graphical means of defining schema details, and it is all tied together through the Micro-soft Repository.

In performance terms, 7.0 has adopted a number of technologies from other high-performance products. Row level locking, improved use of parallel architectures, improved memory allocation and SQL extensions aimed at data warehouse support are some. To prove its claims of scalability to terabyte level, Microsoft has made much of its ?Terraserver? web site. This site is built on SQL Server 7.0 and serves up satellite images covering the world. The compressed data is said to occupy over one terabyte.

Microsoft claims that Terraserver is the largest web database in the world. That may be so, but the performance is less than impressive, and the transactions are hardly representative of a typical database application.

More impressive was the record set by Sybase during the 1998 World Cup with a peak hit rate of 235,356 hits per minute and a total of more than 1.1 billion hits during a month in the tournament. If Microsoft really wants to play with the big boys, these are the numbers to aim for.

The most controversial addition to SQL Server 7.0 is the Decision Support Server (DSS). DSS features include logical and partitioned cubes, drill-through to underlying relational data, and client-side caching. Current Micro-soft front-end analysis tools don?t yet make best use of DSS, but a new release of Excel due in early 1999 will feature extended PivotTable functionality. PivotTable connects client-side tools and the DSS server, using client-side caching to improve overall performance.

To complete its data warehouse capability, version 7.0 includes a set of data transformation services managed via simple wizards, instead of the under-lying Olap technology.

There are many single purpose products able to score over 7.0 in specific areas, but Microsoft doesn?t really need to be best at everything. It is enough to provide the most cost-effective software for the majority of the market. Using the 80/20 rule, if Microsoft can dominate the 80%, the 20% can be picked off later. No effort has been made to allow SQL Server 7.0 to store objects or user-defined data types, as Microsoft believes that Windows NT should provide the capabilities for managing exotic data types.

It does, however, emphasise the one major constraint on SQL Server ? it can never be more successful than NT.

Given the continued slippage of NT 5.0, the surge in interest in Linux (supported by all of Microsoft?s database competitors) and other recent Unix initiatives, it is not certain that NT will become the leading server operating system.

It is reasonable to expect Microsoft?s SQL Server 7.0 to become the volume leading database management systems on NT, but predicting the size of that market is another game.

? Rob Hailstone is research director at Bloor Research