Welcome Guest. | Log In| Register | Membership Benefits

AuthorITies: Redmond Watch


April 12, 1999

The Long Wait: Microsoft's Windows 2000 Missteps

By Stuart J. Johnston

M icrosoft late last month announced a major reorganization aimed at focusing products more closely on users' needs. Among the changes, the company has created five divisions, including the business and enterprise division which will focus almost exclusively on Windows 2000. Long-term, that may benefit enterprise customers in getting the functions they need in the operating system.

Unfortunately, for those waiting impatiently for Windows 2000, the reorganization will have little impact on speeding delivery of the long-awaited major upgrade to Windows NT.

In the near term, there are two new developments that IT decision makers and administrators should know about. First, in its quest to comply with users' No. 1 priority--greatly increased reliability--Microsoft is going to break a lot of existing applications that violate guidelines that the vendor has been trying to promote for the past few years.

I had received a tip that there were going to be compatibility problems with beta three back in late January but it wasn't until mid-March that I could get verification that there is indeed a problem. In calling around, no one I talked to had seen the problem, and it turns out that the technology that enforces the guidelines is only getting turned on for the first time when beta three ships later this month. The first anyone heard about this officially was on a single slide presented at a "local media day" at Microsoft in mid-March.

Microsoft has made significant progress in fixing its end of the problem in the past two months. Still, there will be some applications compatibility problems in beta three. According to one Microsoft product manager, about 20% of today's top 500 commercial desktop applications have problems, ranging from some functions that may not work properly to some applications that may not work at all.

To be fair, this is not just Microsoft's fault. In many respects, the company is guilty of being indulgent of independent software vendors. Many developers have used some sloppy programming methods as well as quick and dirty hacks to make their products work.

The offenses range from writing customized versions of shared system files which an application's installation program copies into system directories, overwriting existing system files, and perhaps breaking some other vendors' applications at the same time.

In the past, particularly with Windows 3.x and 9x, Microsoft sacrificed stability in the operating system in order to provide as much backwards compatibility with existing applications as possible so as to not require users to replace a large number of their applications in an upgrade cycle. Microsoft was a bit more stringent and a bit less backwards-compatible with Windows NT, but stability was still an issue.

The company is very aware of that and, at a Microsoft Business Applications conference in Las Vegas last September, company president Steve Ballmer publicly apologized that NT 4.0 wasn't as stable as NT 3.51. He assured customers that Windows 2000 will be much more stable than current versions of NT.

Surveys conducted by InformationWeek Research have repeatedly verified that stability and reliability are at the top of customers' lists. For example, our most recent survey, conducted in mid-January, found that by far, the primary criteria for evaluating Windows 2000 on both the desktop and the server will be stability. There was a 15- point difference between stability and backwards compatibility as the top criteria for evaluating Windows 2000.

Microsoft has obviously heard that message from users loud and clear.

Unfortunately, however, because Microsoft didn't write the applications, it can only do so much to alleviate the problems. Indeed, company officials in follow-up interviews have admitted that they may not be able to fix all of the problems with existing apps by the time Windows 2000 finally ships.

So, worst case, prepare for what a friend of mine refers to as "another gutwrenching Windows upgrade." At minimum, depending on whether you've got applications that did violate the guidelines, you should be planning on some added staff expenses. You can also begin doing some planning now by evaluating your supported applications to see if they violate the guidelines for "well-behaved" applications.

For a look at the guidelines, go to: http://msdn.microsoft.com/developer/winlogo/win2000.htm

Meanwhile, Microsoft developers have also apparently met further delays in bringing the company's vision of multi-server clustering to reality. A little over a year ago, chairman Bill Gates said that a year hence, the company would make available 16-server clustering. Asked a few days later to elaborate on Gates' proclamation, product managers said that Bill was actually referring to the beginning of the beta test.

So here it is a year later. Where's the beta? Last week at the WinHec conference in Los Angeles, Microsoft demonstrated four-server clustering. But there's still no word on when the beta will actually start, besides the fact that four-way clustering under Microsoft Cluster Services (MSCS) is a long ways from 16-way clustering. When asked about Gates' statements regarding 16-way clustering, a Microsoft official said that Bill was not fibbing--that Windows 2000's kernel is architected to support 16 nodes in a cluster.

Microsoft officials have said for more than a year that multiserver clustering beyond tying a pair of servers together would have to wait until after Windows 2000 ships. But Windows 2000 has slipped repeatedly, so multi-node clustering under MSCS has also slipped. It's still expected to be delivered following shipment of the first three versions of Windows 2000.

Windows 2000 Datacenter Server, the high-end version of the operating system that will support up to 32 processors in a single box and up to 64 Gbytes of RAM, will ship three months after the first three versions. But it will only support four-server clustering in MSCS. Since it's scheduled to ship two to three months after the first three versions-- the desktop and two smaller server versions--it is likely to come sometime in the second or third quarter of 2000.

The hang-up, according to Microsoft officials, has to do with the commodity nature of NT and Windows 2000, as well as the fact that there has been such a dizzying proliferation of hardware interconnectivity technologies. The trick is in testing all the possible configurations and certifying them to work.

As the number of machines that are clustered together doubles, the level of complexity increases geometrically-- and with it the complexity of testing and certifying systems to work correctly with the software.

Perhaps that's why the vast majority of clusters today consist of four or fewer machines. Thus, IT managers aren't likely to scream bloody murder over this latest bump in the road.

In both cases--the issue of backwards compatibility and of four-server clustering instead of 16-server clustering--the negative impact may not be huge. But if Microsoft wants to improve its level of trust with the IT community, the company would be better served if officials were more forthcoming about problems earlier on. Microsoft's refusal to confirm the details of problems until the company is forced to do so doesn't breed the kind of trust that IT professionals demand if they're going to put their business-critical applications on Windows 2000.


AuthorITies Archive

Send Us Your Feedback

Top of the Page

Karyl Scott:
Enterprise View
Karyl will explore the business and technology issues surrounding enterprise systems. 		Charles Pelon:
Eye On IT
Charles explores IT management issues and strategies that business and technology managers face. 		Jason Levitt:
Internet Zone
Jason focuses on the strange, egregious, and the standard technologies of the intranet/Internet. 		Sean Gallagher:
The Bleeding Edge
From his vantage point of managing editor of InformationWeeek Labs, Sean will explore the impact of new technologies on the evolving world of electronic business.