Guild Companies

It may have taken three years to develop, and may take a quarter of a billion dollars in marketing to prime the pump, but the new "Whistler" release of the Microsoft Windows server platform, which was launched on April 24 as Windows Server 2003, is going to be something that tens of millions of companies worldwide are going to be living with for a lot longer than a few years.

One might go as far as calling the new Windows something more appropriate, like Perpetual Windows, if it succeeds like Microsoft expects it to, because once customers buy it, they will be loathe to upgrade to something else. Success has a funny way of killing itself. Just ask IBM about how often companies running its rock-solid AS/400 and iSeries servers have to upgrade. If their workloads aren't growing, they forget they have the machines, doing silly things like sheet-rocking over them in an office or misplacing them in the back of a warehouse because no one has serviced the machines in so long. Unix machines from Hewlett-Packard, Sun Microsystems, and IBM have similar reliabilities and urban legends, and their customer bases are just as happily stubborn about not messing with machines that work. Microsoft, having finally grown up to deliver a real operating system--not the "toy" that Windows NT designer Dave Cutler didn't want to waste his time with--will now discover that success is just as difficult as not succeeding. Success, though difficult to achieve, is a lot more profitable than not succeeding, which is why anyone would bother to try to do a job right. But that doesn't make it easy.

Windows Server 2003 has had four names and has been delayed three times, and now six different editions (the Web, Standard, Enterprise, and Datacenter editions for 32-bit Pentium machines and the Enterprise and Datacenter editions for 64-bit Itanium) are finally available, in 13 different languages, with the remaining languages rolling out over the next 30 to 45 days. Windows 2003 is a product that has been bent like a pretzel and stretched like silly putty to cover the gamut of IT uses. The Web Edition is a $399 defense against Linux and Unix for basic infrastructure jobs. The Standard and Enterprise Editions aim to cut off Linux at the pass for real-world computing, to give customers with Windows 2000 Server and Advanced Server an upgrade path, and to extend Windows into 64-bit computing (Enterprise), where a lot of the action in the midrange is. The 32- and 64-bit Datacenter Editions aim to unseat Unix as the preferred industrial-strength operating system for commercial data processing, and it most certainly has the goods to compete with big iron and to make the Unix vendors wonder how they are going to compete against big Intel boxes. It's a good question, and it's also a good thing that a lot of applications are not yet ready to run on Windows 2003, including many of Microsoft's own products. The Unix vendors have some time to rejigger their pricing models to meet the Windows Server 2003 threat, or to port their Unixes to Intel's Itanium or Advanced Micro Devices' Opteron and be done with it, like HP has done.

It's tough out there, and the server platform vendors are talking about how their latest product allows people to do more with less. Steve Ballmer, Microsoft's CEO, echoed that sentiment at the Windows Server 2003 launch event in San Francisco last Thursday, and even kicked off a Microsoft marketing campaign that promises that Whistler will allow companies to do more with less. While Windows Server 2003 is going to appeal to the 4.5 million Windows NT 4.0 shops in the world, which comprise about 35 percent of Windows servers installed worldwide--and it will be appealing to Windows 2000 customers who want something that is better, stronger, and faster--there's only one problem: Being able to do more with less is still not as inexpensive (from a budgetary viewpoint) as doing nothing at all.

Back in the late 1990s, when money was loose and times were good, Microsoft had an easy time selling Windows NT, especially after Version 4.0, arguably the first usable piece of server code Microsoft had ever delivered. Windows 2000 was set to come out just before the Y2K freeze hit in the middle of 1999, and Microsoft pushed it out to February 2000, just as the dot-com bubble burst and took down the stock market with it. Microsoft could have gotten Whistler out the door in 2001 or 2002, and it wouldn't have mattered. No one would have bought it any more than they had to as part of new equipment purchases or if it was required by application software. And now, in 2003, that has not changed. Those millions of Windows NT 4.0 shops are certainly in need of support, which they cannot get anymore, and they will be more inclined to jump to Windows 2003 than Windows 2000 because the prices are the same. But mark my words: For many millions of shops, the next Windows operating system they buy will be five years from now. Microsoft is entering the old 80-20 game, where 80 percent of the customers drive 20 percent of sales and 20 percent of customers drive 80 percent of sales. Installed bases get calcified as the software becomes successful, and only the risk of security threats to systems will compel companies to upgrade. (If hackers didn't exist, they might have to be invented, I presume, to coax companies along the upgrade path.)

To be sure, Microsoft is going to enjoy a very healthy upgrade business--even with its Licensing 6.0 pricing model, which has raised prices for many Windows shops--and it is going to love the big-iron business it gets, because those Datacenter Edition licenses are going to bring it hundreds of thousands of dollars of nearly pure profit as companies adopt the Microsoft middleware stack to run some of their applications. Unix did it to proprietary mainframes and minicomputers in the early 1990s, and Windows is going to do it to Unix in the early 2000s. However, Microsoft would do well not to celebrate too much. Just like platform vendors a decade earlier tried to dismiss Windows NT as a toy, Microsoft's dismissing Linux as a toy is utter foolishness. Linux is going to catch up to Windows Server 2003 a lot faster than Windows caught up with Linux. Microsoft has a very short window of opportunity, and it had better make the best of it. Because it is going to get a lot harder in 2005, when Linux can scale as far as anyone needs it to and can run a lot of the same applications.

In the meantime, Microsoft is going to have to use some of that $250 million it has earmarked for promoting Windows Server 2003 to quantify the performance and reliability benefits that Windows NT and Windows 2000 shops can expect from an upgrade to Windows Server 2003, and it will similarly need to flesh out the real savings that customers can expect from the new operating system. While Microsoft has been precise in its claims--an eight-fold increase of Windows reliability compared with Windows NT 4.0 seven years ago, a 30 percent increase in the efficiency with which IT Infrastructures run, and a 20 to 30 percent reduction in the number of physical servers needed to perform certain workloads, just to name three stats Microsoft is throwing around--the source of these claims is vague and unsubstantiated. The TPC-C online transaction performance benchmarks that Microsoft has delivered with its partners are stunning, to be sure. But this is not enough.

Another thing that Microsoft needs to do is to assure customers who are running applications on Windows NT 4.0 and Windows 2000 that their applications can run on Windows Server 2003. There are some applications--particularly those that rely on the Internet Information Services 5.0 Web server--that will not run on the new operating system without some major changes, and Microsoft had better help customers identify these applications and help them do the ports before they trip on the problems themselves. Customers can start by downloading the new Windows 2000 Resource Kit. The company is also giving away the Enterprise Edition of Windows Server 2003 as part of an evaluation kit on CD and as a download on a trial basis, until July 31. Don't buy anything until you've played around with Windows Server 2003 for a while. Take your time, and do it right.