The long road to Yukon

By Mike Ricciuti
Posted on ZDNet News: Dec 23, 2003 12:00:00 PM

But Microsoft executives dispute that notion. They told CNET News.com that the delivery of Yukon--which marks the debut of the company's key unified storage technology--won't affect other future products.

The company also for the first time gave insight into the difficulty of building the new storage technology--a pet project of Microsoft Chairman Bill Gates for more than a decade and one of the most ambitious and time-consuming projects the company has ever attempted.

"Microsoft's dream is to have this unified single storage technology that we can use across products," Stan Sorensen, a product manager at the company, told CNET News.com. "As long as I have been involved in servers--more than 10 years at Microsoft--we have been trying to achieve this."

Yukon, an update to Microsoft's SQL Server database, will usher in the storage concept. The storage technology has become a massive undertaking that will ultimately affect nearly all of the company's key products.

Microsoft said the storage technology helps to blur differences between data types and will make it far easier for people to search for and find documents, e-mail messages and multimedia files scattered across their hard disks and on networked computers.

In addition to Yukon and Longhorn, a new version of Microsoft's Exchange messaging server, code-named Kodiak uses the storage technology. All three are expected to debut in 2006. Microsoft hasn't said exactly which other new products will use the storage technology. "So many products at Microsoft use SQL Server, it's hard to know how many dependencies will exist," said Chris Alliegro, an analyst with Directions on Microsoft.

Clearing up some of the mystery surrounding Microsoft's efforts, Sorensen said that products using the storage technology will not have common code, as many industry watchers have thought. Instead, individual product groups within Microsoft will each develop storage systems based on a common design specification.

Yukon and Longhorn "are two completely separate projects--different development teams, different development schedules--so there is no schedule impact. Yukon delays will have no bearing on Longhorn," Sorensen said. "There is a shared philosophy, not so much shared code."

For instance, Yukon, Longhorn and Kodiak will use "a common data schema, a common way to build relationships between data, and a common query method," he said. There is "some very low-level code that is shared, but that has been (completed) for years."

Early speculation was that Longhorn's new file system, called WinFS, would actually include a copy of the SQL Server database engine.

Microsoft's own descriptions of its storage work may have added to the confusion. "There is no coincidence that as we are developing this Yukon version of SQL Server, we thought, 'Wow, we can take some of that and have it be in the (Longhorn) file system,'" Gates told CNET News.com in October.

"I think a lot of people have the impression that Yukon is buried in WinFS, and Microsoft is starting to take pains to remove that impression from people's minds," Alliegro said.

All eyes on Yukon
Microsoft wants to guarantee the reliability and security of the storage technology before its high-profile debut in Yukon. Even if products do not share code, a significant delay in building the new software due to architectural problems could still derail other product schedules. Yukon is, in essence, a proof-of-concept for the entire storage effort.

"The Yukon delay does cascade into other products to a certain extent," RedMonk analyst Steven O'Grady said. "Yukon is very important to other product releases, and to Longhorn, it's crucial."

The storage concept has been well received by customers and analysts alike. But delivering working code has been a challenge. Yukon was originally slated for delivery this year. Last summer, Microsoft said it had delayed Yukon's debut until late next year.

Microsoft gave more details of Longhorn in October, and it has distributed more than 12,000 copies of an early test version of Yukon to software developers over the past few months. But a wide-scale test version of Yukon, expected earlier this year, won't arrive until the first half of next year, according to Sorensen.

Among the early testers is Microsoft itself, which will put Yukon through its paces on some of its internal systems.

"Come hell or high water, (Microsoft) wants to be solid on that beta" early next year, Alliegro said.

The main reason for Yukon's delay is the overall complexity of building the software, Sorensen said. "It's just a big release, the biggest SQL Server release (in more than five years). There are lots of moving parts," he said.

Adding to the delay is a renewed focus on ensuring Yukon's reliability and security--an offshoot of Microsoft's 2-year-old Trustworthy Computing initiative.

A "bulletproof" launch
With Yukon, Microsoft is aiming to take away enterprise database sales from market leaders Oracle and IBM. That's added to the pressure on developers to make this release "bulletproof," according to analysts.

Microsoft has redoubled testing efforts and will shut down Yukon development for five to 10 days in the coming months to do a "security scrub," in which engineers pore over millions of lines of SQL Server code. "We're making sure we really nail the security bits," Sorensen said.

Another area of difficulty for Microsoft's engineers has been integrating new software development technology, called the Common Language Runtime (CLR), into Yukon while simultaneously developing new database components using the CLR.

"We made a big bet on the CLR, and that is causing a lot of work on development and testing to integrate the CLR into the database," Sorensen said. "And we are building components on the CLR itself. So it is taking a number of additional cycles to do that, and there is a lot of rewriting we are doing in the product (Yukon)."

In Yukon, the CLR will allow Microsoft customers to build applications in the database using a variety of programming languages in addition to SQL Server's native language, called Transact-SQL. Developers also will be able to build systems that use more familiar languages, such as Visual Basic and Visual C#, Sorensen said.

Incorporating the CLR into Yukon means the millions of programmers already trained in Visual Basic will be able to write SQL Server applications, potentially broadening the appeal of the product.

In addition to the CLR, Yukon will include better support for Microsoft's .Net programming software, so Web services code is easier to build and launch on the database.

The delay in delivering Yukon most likely won't cause too many problems for Microsoft's database customers, RedMonk's O'Grady said. "I think Microsoft customers are frustrated by product delays, but this is not a huge deal for them. It's not a deal-breaker for them."

Instead, the holdups could affect the company's revenue projections and planned product releases for 2004 and beyond. Microsoft is hoping for a jump in revenue from Yukon sales late next year and in 2005, as customers upgrade their software. And if Longhorn is affected by Yukon's delayed schedule, many other products could be too.

To be sure, Longhorn is still a work in progress and isn't likely to ship before 2006. Some analysts don't expect it to debut until 2008 or even later. But new versions of Office--Microsoft's most profitable product--and other technologies are closely tied to Longhorn's schedule. Gates and other company executives have said that "Longhorn will impact Microsoft development across the board," according to O'Grady.

That puts the pressure squarely on Microsoft to complete Yukon next year, because it will ultimately tie into the company's larger plans for WinFS and Longhorn. Yukon's most noteworthy feature is the new unified storage system, which will support relational data, XML (extensible markup language) files, and user-defined data types.

CNET News.com's Martin LaMonica contributed to this report.