The Charge to SQL95 | Inside Microsoft SQL Server 2000

As late as the beginning of 1994, the SQL Server development team had planned for the next version of the product to pick up the Sybase System 10 source code and new features. But the termination of joint development changed that plan. Sybase contributed no more code, and Microsoft released no System 10 product. Except for a couple of bug fixes, the last code drop received from Sybase was in early 1992 just as version 4.2 for OS/2 was shipping.

Time was at a premium. Besides promoting sales to new customers, Microsoft had to fight for its significant installed base. Sybase was to deliver System 10 for Windows NT later that year. That would be a potentially easy upgrade for Microsoft's installed customer base; so if Microsoft lost customers to System 10, they'd likely lose them forever.

Microsoft quickly planned for an ambitious release that was loaded with new features and performance improvements. It was tagged SQL95, borrowing the working moniker from the well-known upcoming Windows 95 release. Because the Big Question of 1994 was "What are your plans for replication?", replication became a keystone of the release. So did scrollable cursors, a feature that the developers had learned was necessary (from the ill-fated dBASE IV interface) to bridge the impedance mismatch between many record-oriented applications and a relational database. No mainstream DBMS product had yet provided a fully functional implementation of scrollable cursors in a client/server environment, and the SQL Server team believed that it was imperative to add this feature to the database engine. They had also been working on a dramatic new set of management tools, code-named Starfighter (today's SQL Server Enterprise Manager), which would also be included in the next release. The new feature list went on and on.

Microsoft's customers were clamoring to hear about the plans for SQL Server post-Sybase. So on June 14, 1994, Microsoft put together a briefing event in San Francisco for key customers, analysts, and the press. Jim Allchin, Microsoft senior vice president, walked the attendees through the plans for the future and for the SQL95 release. Attendees were impressed with the plans and direction, but many were openly skeptical of Microsoft's ability to deliver such an impressive product by the end of 1995.

NOTE
Some industry press began to sarcastically refer to the planned release as SQL97 and even SQL2000. Internally, the development team was still targeting the first half of 1995. To outsiders, they were more cautious—after all, this is software development, which is still more art than science. Stuff happens, so the SQL Server team said nothing more ambitious than 1995. (Everyone assumed that the planned date was December 31, 1995, and that they'd miss that date, too.) The rampant skepticism only served to motivate the SQL Server team even more to show that they could deliver. After all, the team rightly felt that it had already delivered an impressive release independent of Sybase. But no one gave them credit for that, so they'd just have to show everyone.

The team worked incredibly hard, even by Microsoft standards, to make this deadline and still deliver a full-featured, high-quality product. The first beta was released at the end of October 1994. Although Starfighter was not feature-complete yet, the database server was complete, and because the server takes the longest lead time for beta sites to really stress it, they went ahead with the beta release. This release was followed by a series of beta updates during the next several months, along with gradual beta-site expansion, eventually surpassing 2000 sites.

For nine months, dinner was delivered each night for those on the development team who were working late—usually a majority. On June 14, 1995, the product was released to manufacturing. Microsoft SQL Server 6.0 (SQL95) had shipped within the original internal target date, much sooner than nearly everyone outside the team had expected. It was an immediate hit. Positive reviews appeared in nearly all the trade publications; even more significant, none were negative or even neutral. Even more important than the press reviews was customer reaction, which was terrific.

InfoWorld, in its second annual survey of the 100 companies with the most innovative client/server applications in the previous year, showed Microsoft SQL Server as the number-two database. SQL Server jumped from 15 percent to 18 percent of those surveyed as the database server of choice—for a virtual tie with Oracle, which dropped from 24 percent to 19 percent. Sybase rose from 12 to 14 percent. Three of the top 10 applications highlighted by InfoWorld were built using Microsoft SQL Server.

Of course, the SQL Server development team was happy to see this data, but they took it with a grain of salt. Other data could be interpreted to suggest that Microsoft SQL Server's presence was substantially less than the surveys indicated. And the team recognized that they were still relative newcomers. From a sales perspective, Oracle was clearly king of the hill, and Sybase, Informix, and IBM were also formidable competitors in the DBMS market. Microsoft had not previously been a significant competitive threat. But now it was, and all companies armed their sales forces with bundles of information on tactics for selling against Microsoft SQL Server. Sybase, Informix, and Oracle all promised hot new releases. Although the SQL Server team had been sprinting for nearly four years, now was certainly not the time to get complacent.

Team-Building with Top Talent

Besides working hard on the development of version 6.0, Microsoft was also working to increase the size and strength of the team. It had built a small, crackerjack team that had delivered SQL Server for Windows NT, and this team was the core that delivered SQL95. But the team needed more people, more expertise, and exposure to broader ideas. So they went after top talent in the industry.

Microsoft attracted some industry luminaries—Jim Gray, Dave Lomet, and Phil Bernstein. They also attracted a lot of lesser-known but top development talent from throughout the industry. For example, DEC shopped its Rdb product around, looking to generate cash, and Oracle eventually spent a few hundred million dollars for it. But Microsoft didn't want to buy Rdb. Instead, it hired many of the Rdb project's best developers, augmenting this move by hiring several of the best recent masters' graduates who had specialized in databases.