Building the Billion Dollar Database: Microsoft SQL Server Climbs to New Heights

REDMOND, Wash., June 28, 2001 — In an intensely competitive business world where only the strong survive, what is the ultimate measure of success? Rave reviews in the media? They’re nice. Reports that customers are switching from a rival company’s offering to your product? Not bad.

SQL Server Timeline

How about this: News that your product has reached $1 billion in annual sales? Now that’s a milestone worth celebrating. Ask Steve Murchie he’s Microsoft’s group product manager for SQL Server Marketing, and Microsoft SQL Server has joined the ranks of products that rake in more than $1 billion a year. That puts SQL Server in elite company. After all, not many companies can even boast of total sales greater than $1 billion a year.

“Is hitting $1 billion significant?” Murchie asks. “Absolutely. There just aren’t that many products in this business — whether you’re talking about Microsoft or the software industry as a whole — that have achieved that level of revenue. We’re elated at hitting this milestone.”

Elated, but not surprised. Since the release of SQL Server 7.0 in late 1998, Microsoft has played an increasingly important role in the highly volatile and fast-expanding market for enterprise relational database management systems — databases designed and built to help companies store and utilize massive large amounts of critical corporate data. And when SQL Server 2000 became available last summer, Microsoft cemented its position as a leading vendor of enterprise-class database software.

“It was definitely in our plan to hit $1 billion in this year,” Murchie says. “And we never had any doubts that we would reach that milestone.”

If Murchie sounds confident, a look at the number reveals why. According to estimates from Dataquest, SQL Server revenue stood at about $350 million in 1998. A year later, that figure had jumped to more than $600 million. Last year, SQL Server was fast approaching the $900 million mark. When a product enjoys year-to-year growth rates of nearly 50 percent, it’s hard to blame someone for their confidence.

It wasn’t always that way with the SQL Server group. When Murchie came to Microsoft four years ago to join the SQL Server team, most industry observers would have laughed at the notion that Microsoft SQL Server might someday be a billion-dollar-a-year product.

“When I first started, we were still shipping SQL Server 6.5,” Murchie remembers. “Those were challenging times. We couldn’t quite compete with the leaders in the industry back then on scalability or reliability. Oracle was constantly berating us for missing features, and it was pretty hard to respond. Now we own most of the major benchmarks, we’re the most popular Web database, and we deliver great uptime. The quality of SQL Server speaks for itself.

“But,” he adds, “it was a pretty long road getting here.”

A long road indeed. The story of Microsoft SQL Server stretches back beyond the mid-1990s and the release of SQL Server 6.5. In fact, Microsoft didn’t build the first version of SQL Server, and it didn’t run on Microsoft Windows. It was built by Sybase, and ran on the IBM OS/2 platform. The earliest history of Microsoft SQL Server extends back to 1986, when Microsoft entered into talks with Sybase to discuss licensing DataServer, a database product built to run on UNIX computers.

Those talks led to a product called Ashton-Tate/Microsoft SQL Server 1.0, which shipped in May of 1989. Back then, Ashton-Tate was the unquestioned leader in the PC database world. Ashton-Tate/Microsoft SQL Server 1.0 was an adaptation of Sybase’s DataServer for UNIX, to be marketed by Ashton-Tate and Microsoft for the OS/2 market.

Ashton-Tate/Microsoft SQL Server 1.0 got off to a slow start. OS/2 was not the big seller everyone had expected it to be, and in early 1990, Ashton-Tate was struggling to survive. By the summer of 1990, Microsoft and Sybase had terminated the marketing agreement with Ashton-Tate and were ready to ship a new version of the product, simply called Microsoft SQL Server 1.1. In addition to a number of bug fixes, version 1.1 included an important new feature — support for Windows 3.0, which had shipped in May of 1990.

In those days, Microsoft played a limited role in the development of SQL Server. As described in the book ” Inside Microsoft SQL Server 2000 ,” Microsoft provided client software and utilities, programming libraries and administration tools in SQL Server 1.1. The core SQL Server engine, however, was still produced entirely by Sybase. Microsoft developers didn’t even have access to the source code, and any requests for changes — even simple bug fixes — had to go to the SQL Server team at Sybase.

Even so, says Stan Sorensen, director .NET Enterprise Server marketing at Microsoft, the relationship with Sybase was critical. “That partnership launched us into the market,” he says, “and it gave us a code base and a product line that we could build on.”

Changes in the Industry

In early 1991, a couple of important factors were working together to shake up the computer industry — OS/2 was a failure and Windows 3.0 was a runaway success. By May, Microsoft and IBM announced that they would no longer work together on the development of OS/2. Instead, Microsoft was devoting its energies to a new operating system, code-named “NT.” Meanwhile, the SQL Server teams at Microsoft and Sybase were working on a new version of Microsoft SQL Server, to be called SQL Server 4.2, which would put it in sync with the Sybase’s latest DataServer release for UNIX, version 4.2. By then, Microsoft’s SQL Server developers were on much more equal footing with their counterparts at Sybase. Released in May of 1992, the new 32-bit version of SQL Server was built to run on OS/2 2.0 — which had yet to ship — as well as Windows 3.0.

Developers at Microsoft immediately switched their attention to a version of SQL Server that would be compatible with the upcoming release of Microsoft Windows NT. Meanwhile, developers at Sybase were focused on creating an entirely new version of their DataServer, called System 10. The new release of Microsoft SQL Server, which shipped in August 1993, a month after the release of Microsoft Windows NT 3.1, was called Microsoft SQL Server for Windows NT. Although it had been substantially rewritten to work smoothly with Windows NT and offered some enhanced management and networking features, the new edition was essentially identical to SQL Server 4.2 — the goal had been to provide a feature-complete version of SQL Server for the new operating system.

SQL Server for NT was a significant success. Tight integration with Windows NT made SQL Server easier to use than any high-end database ever produced, which meant that for the first time, Microsoft had a database that could begin to compete with products designed to run on UNIX. By the end of 1993, the relationship between Microsoft and Sybase was becoming increasingly problematic as the two companies found themselves competing for the same customers. In April of 1994, the five-year-old joint development agreement between the two companies was terminated.

In the wake of the end of the Microsoft-Sybase partnership, the Microsoft SQL development team — which had grown to include more than 50 developers — shifted attention to an ambitious new release that would offer significant improvements in both features and functionality. By most accounts, the results were promising. “Microsoft may finally be gaining credibility as a supplier of enterprise-capable database technology,” reported Information Week in October of 1995, three months after Microsoft SQL Server 6.0 hit the market. Less than 10 months later, SQL Server 6.5 was released to manufacturing.

A Rapidly Growing Market and a New Beginning

That was a key period for both the Microsoft SQL Server team and the database market as a whole. “SQL Server 6.5 just didn’t do that well on key benchmarks,” Sorensen remembers. “For example, we just couldn’t offer the scalability and reliability needed to handle a medium- to large-sized company’s entire human resources system. So we took a long hard look at the market, stopped what we were doing, and decided to start all over again.”

The market they looked at — database sales on the Windows NT platform — was in the midst of explosive growth. Worth just $180 million total in 1995, it had jumped to $1.5 billion a year later, according to Dataquest. (The UNIX market had grown from $2 billion to $2.5 billion.) Overall, Oracle was king of the hill, controlling 25.5 percent of database sales, with IBM second at 22.5 percent and Microsoft third, with 11.5 percent.

Step one was a concerted effort to recruit the world’s leading database development talent. The results were spectacular. Jim Gray, Dave Lomet, and Phil Bernstein — three true pioneers in the field — joined the group. Meanwhile, when DEC sold its database product to Oracle, Microsoft hired many of DEC’s top database developers.

The new team had one overarching goal as they mapped out the next version of SQL Server — to entirely re-architect the database engine so that it would scale as much as the users wanted it to. That meant building in the capability to take advantage of the ever-faster processors and ever-increasing memory capacity of modern servers. It also meant adding a host of critical new features, including full row-level locking and a new query processor that could efficiently handle
“ad hoc”
queries — a requirement when dealing with data warehouses and Internet-based applications.

It took almost three years, but when Microsoft SQL Server 7.0 was announced at COMDEX on November 16, 1998, more than a dozen major corporations — including Pennzoil, HarperCollins,, and CBS SportsLine — were already in full production with a wide range of mission-critical enterprise applications based on the new product.

“The release of SQL Server 7.0 was a big turning point,” Murchie says. “We touched every part of the product, and added critical features like row-level locking and new levels of self-tuning and self-management. We also built in OLAP capabilities. A huge number of independent software vendors signed up to work with us as a result, many of whom hadn’t looked at SQL Server 6.5 because of some of its limits.”

SQL Server 7.0 quickly earned rave reviews in the press. “SQL Server is a part of the platform that makes data warehouse and business intelligence much more accessible to customers,” reported VarBusiness . “SQL Server 7.0 out of the box contains a robust set of widely supported services and technologies unmatched by Oracle8i, Oracle Data Mart Suite or Oracle Warehouse Builder.”

“There’s no doubt that SQL Server 7.0 is faster and more scalable,” added Information Week .

The results showed up almost immediately in SQL Server’s sales figures. According to the Giga Information Group, Microsoft SQL Server 7.0 enjoyed an astonishing 60 percent sales increase in 1999 compared with 1998. Oracle’s growth on the Windows NT platform, meanwhile, slipped from 47 percent in 1998 to just 15 percent in 1999. By the end of 1999, Microsoft had whittled Oracle’s lead in the NT market to a narrow three percent, with Oracle controlling 39 percent of the database market for Windows NT (down from 46 percent in the year before), while Microsoft clocked in at 36 percent.

The March to $1 Billion

Success continued to mount in 2000. Sales of SQL Server for the year reached $894 million, according to DataQuest, up from $618 million in 1999, and SQL Server 7.0 began to obtain key benchmarks. More and more major enterprises were finding that SQL Server was not only significantly less costly than products offered by any other vendor, but was also more powerful and more reliable.

Meanwhile, work on the next version of SQL Server was moving forward. Originally intended as an upgrade of SQL Server 7.0, the decision was made to go ahead with a full-blown major release that would offer important new features, including native XML capabilities, indexed views and support for distributed partitioned views. The development team announced that they were aiming for at least a 20 percent increase in performance in every measurable area of the product.

In February 2000, a beta version of the new edition, eventually known as SQL Server 2000, went through a series benchmark tests. Among the results was a set of records on the important Transaction Processing Performance Council’s TPC-C benchmark, which demonstrated clearly that Microsoft SQL Server 2000 offered the industry’s best price-to-performance ratio on clustered hardware. At the same time, SQL Server 2000 recorded a 48 percent improvement on the SAP R/3 Sales and Distribution benchmark.

When the finished version SQL Server was ready for release on August 7, 2000, it was already clear that SQL Server 2000 was going to be a success. According to an announcement from Microsoft, more than 100 important enterprise customers were prepared to go live with SQL Server 2000 within three months of the initial release date. In the months that followed, leading companies ranging from Credit Suisse First Boston to to announced that they were switching to SQL Server 2000. In February, SQL Server 2000 was named the Datamation Product of the Year for Data Warehousing and Business Intelligence. More recently, Microsoft announced that SQL Server is running Verizon’s nine-terabyte ordering, billing and repair system.

The success of SQL Server 2000 paved the way for the next two important milestones: leadership in the Windows database market and $1 billion in annual sales. The first milestone was achieved before the end of 2000, according to Gartner Dataquest, which reported on May 23 that SQL Server accounted for 38 percent of new database license sales on the Windows Server platform that year, compared with 37 percent for Oracle.

That, coupled with the runaway growth rate of SQL Server, which saw sales increase more than 45 percent in 2000, gave Steve Murchie confidence that SQL Server 2000 would hit $1 billion in sales this year.

“SQL Server has always been the leader for ease-of-use, support and manageability,” Murchie says. “Now we’ve built in all of the enterprise capabilities that are essential for companies running big transaction systems and massive data warehousing applications. It’s paid off. Today we’re the number one database in the NT market.”

“We’ve succeeded because we never wavered from our most important goal — to do right by our customers,” Stan Sorensen adds. “We never got distracted by trying to chase after the latest and greatest market, which is what happened to our competitors. That’s why their sales are essentially flat while we’re continuing to grow, and grow big.”

So what do you do for your next trick if you manage a product that’s gone from zero to $1 billion in less than 10 years?

“If you think about it, it’s really only been six or seven years since we took over development of SQL Server from Sybase, so that’s pretty tremendous growth,” Murchie says. “And we’re excited about our next version, which is codenamed Yukon, because of the opportunities it offers for expanding the scope and value of databases. But for the moment, we’re pleased to let the success of SQL Server 2000 speak for itself. I’m happy that we can just keep our heads down and quietly go about the work of continuing to build the world’s best database product.”

Related Posts