Chapter 1: Architecture and Installation

To understand the power and behavior of SQL Server 2000, it's helpful to look at its roots. This chapter focuses on the history and architecture of SQL Server, which will give you a preview of the product's personality.

SQL Server's Roots

Microsoft SQL Server (actually called Ashton-Tate/Microsoft SQL Server at first) was born from a deal forged by Microsoft and Sybase on March 27, 1987. The deal gave Microsoft the exclusive rights to Sybase's DataServer product for OS/2 and provided Sybase with a large base of OS/2 customers. This gave Sybase credibility in the market, as well as royalties from any sales Microsoft made.

When Microsoft SQL Server 1.0 shipped, it was available only for OS/2. It was not until version 1.1 appeared that it supported Microsoft Windows 3.0; however, that support was limited to the client tools. When this release shipped, Microsoft did not see SQL Server as a revenue-generating product but as an incentive for customers to buy OS/2 LAN Manager.

SQL Server 4.2 was released in March 1992 as the next joint effort between Sybase and Microsoft. It was the first server release that supported Windows NT. In this release, minor GUI administration tools were added to help DBAs. If you were to look at this release now, you'd laugh at its lack of administration GUI tools compared to today's tools and wizards.

As it turned out, the code that was exchanged in this release of SQL Server would be the last venture between Microsoft and Sybase. By March 1994, the database world had changed dramatically. Sybase no longer needed Microsoft to boost its credibility. Microsoft felt it didn't need Sybase's database expertise any more. The ties were severed.

After the relationship was terminated, Microsoft needed to put together a development team quickly, in order to release the next generation of SQL Server. Microsoft's role had gradually increased from support and maintenance to codevelopment, and an entire team had to be built for their first major release without the assistance of Sybase.

That occurred in June 1995: SQL Server 6.0 (originally code-named SQL95) focused on replication and usability features. It also introduced Enterprise Manager- at the time, no other vendor had a similar management system. Less than a year later, SQL Server 6.5 was released to fix many of the bugs and usability problems in 6.0. Both of the 6.x releases were available for Windows NT.

Even before SQL Server 6.5 was released, a separate team was brought together to work on the next release, code-named Sphinx. Sphinx was a complete rewrite of the database engine and Query Optimizer. This release addressed some of the complaints about SQL Server's lack of row-level locking capabilities and storage engine problems, namely SQL Server's use of devices. On December 2, 1998, SQL Server 7.0 was released to the market. This release could really be considered version 1.0 in Microsoft's new strategy.

Much internal discussion went on at Microsoft whether to name the next release 7.5, 8.0, or stick with the Windows naming convention of 2000. As the planning teams began to gather a wish list for the next release, they decided there were enough features to consider it a major release and followed the company naming standard, calling it SQL Server 2000. If you look at the internal name for SQL Server, however, you'll find that the base version of SQL Server 2000 is actually called 8.00.194. This release focuses on scalability and reliability. It's clear this release is an attempt to attack Oracle9i's stronghold in the scalability market.

Microsoft has big plans for future releases. The next release, code-named 'Yukon,' again focuses on scalability and security. This focus is leading to the further development of the shared-nothing approach, which is covered later in this book. Essentially, this model allows you to scale a very large database out to any number of servers and make the database appear as if it were running on one server.