[转贴]CIO Update: How to Select a Client Operating System

小样

CIO Update: How to Select a Client Operating System  
  
5 May 2004

Michael A. Silver   

Selecting a client operating system (OS) is often confusing. Most enterprises select Windows, but selecting the wrong version can result in increased costs and risk. And then there's Linux.













*
Recipients who do not own this document will receive a summary only.











Browse Topics


Advanced Search










Ask an Analyst




Rate This Document








Downloads:



cio_update_how_.pdf (33.4KB)









Analysis



Selecting a client operating system (OS) is often confusing. Most enterprises select Windows, but selecting the wrong version can result in increased costs and risk. And then there’s Linux.

Windows runs on the vast majority of client PCs worldwide. However, selecting which client OS to deploy remains a complex, confusing decision. Many enterprises have PCs that run legacy versions of Windows, and they must bring in newer versions of Windows.

How much longer can I run Windows NT Workstation v.4 and Windows 98?

You should be well on your way in moving off Windows 95, Windows 98 and Windows NT Workstation v.4 (NTW4). Critical users, such as those with significant monetary or legal exposures — for example, users covered by Health Insurance Portability and Accountability Act (HIPAA) requirements — should be off a client OS by when Microsoft ends assisted its support:

Windows 95, 31 December 2001
NTW4, 30 June 2003.
Windows 98 support was also supposed to end on 30 June 2003, but has been extended. All users should be migrated off Windows 98 and NTW4 by year-end 2004. Windows 95 users should have been migrated off that platform by mid-2003, based on Gartner’s recommendations.

Microsoft has extended its support for some systems: It has stated that it would provide fixes for severe security holes on NTW4 through 30 June 2004 (it has sold custom support contracts for NTW4 through 2005). It will extend this support and supply security patches for NTW4 through at least year-end 2004 (0.7 probability). Microsoft has extended support on Windows 98, and it will have fee-based assisted support and security fix support through 30 June 2006. Gartner believes that Microsoft is offering this support primarily to appease emerging markets that may be interested in Linux as an alternative to Windows (see “Microsoft Extends Windows Support to Prevent Defections to Linux,” FT-21-9810).

Mainstream enterprises that purchase new hardware and software increasingly will find Windows 98 support lacking. Thus, Gartner’s guidelines for when to migrate all users off NTW4 and Windows 98 haven’t changed. The longer that you run older, unsupported OSs, the more you risk problems because of unresolved security or other issues, and the more likely you will encounter problems with running applications, printers, PCs or other devices.

Which version of Windows should I deploy?

If you haven’t spent significant resources on or deployed Windows 2000, skip it and deploy Windows XP Professional instead. However, an impediment to Windows XP is a lack of support from some smaller independent software vendors (ISVs). In fact, some ISVs have only recently added support for Windows 2000, which is more than four years old. Research the applications that you run to ensure they are supported on the OS to which you plan to move.

The main reason to choose Windows XP over Windows 2000 is the support life cycle, not new features. Although Windows 2000 still “feels” relatively new, it is in the middle of its life cycle. As of 31 March 2004, Microsoft stopped licensing Windows 2000 directly to hardware original equipment manufacturers (OEMs). Some OEMs will stop shipping it on new PCs, while others will preinstall Windows 2000 using downgrade rights. If you use downgrade rights to run Windows 2000 Professional on new PCs, understand that Microsoft and its PC OEMs will not be responsible for support for the downgraded OS.

On 31 March 2005, Microsoft will likely move Windows 2000 into the extended phase of its life cycle. At that point, assisted support will be fee-based, and new bugs will be fixed and given only to the enterprises that pay to create the fixes. However, Microsoft will extend its critical, security-related patch support for all enterprises past Windows 2000’s end of life on 31 March 2007 (0.7 probability). “Keep Windows 2000 Longer, but Don’t Skip XP” (TG-18-5701) and “Microsoft’s New Client-OS Support Plan: Recommendations” (TG-18-5700) discuss Microsoft’s support life cycle and Gartner’s recommendations.

Most enterprises that have deployed Windows 2000 should have already begun to deploy Windows XP on new PCs. However, in most cases, don’t upgrade PCs running Windows 2000 to Windows XP because there often is little return on investment.

In choosing a Microsoft client OS, consider Microsoft’s support as well as future plans. In the second half of 2004, Microsoft will ship a significant service pack for Windows XP that will attempt to prevent hackers from exploiting security holes (0.7 probability). This service pack will likely be effective when deployed on hardware that supports the No eXecute instruction, which is on Advanced Micro Devices (AMD) Athlon64-based PCs, and will be on Intel-based desktops by year-end 2004 and Intel-based notebooks in 2005. This service pack will compel enterprises to adopt Windows XP, at least on new hardware.

Microsoft will not ship Windows Longhorn before 2007 (0.7 probability) and will ship an interim version of Windows in 2005 (0.6 probability) (see “Will Longhorn Really Be the Next Version of Windows?” SPA-21-8913). However, Microsoft hasn’t committed to the interim release, so don’t plan your deployments around it. Most enterprises will not be ready to begin mainstream deployments of Longhorn until 18 months after it ships (0.8 probability), likely in 2008 or 2009.

While waiting for Longhorn, you have three choices:

Skip Windows XP to maintain a homogeneous Windows 2000 environment.
– That may require you to push Longhorn PCs out before you’re ready — and perhaps before you have the necessary application support — because Windows 2000 support will end in early 2007.

– Critical users should be migrated before then (and mainstream users within six months after support ends).

Deploy Windows XP to all PCs (new and legacy) to fill the gap between Windows 2000 and Longhorn.
Adopt Windows XP on new PCs.
– This will eliminate most of your Windows 2000 installed base through hardware attrition before Windows 2000 support ends, and you can run Windows XP until you’re ready to deploy Longhorn.

The level of change in Longhorn may prompt you to plan relatively fast deployments regardless of what you choose to do in the meantime. The question is determining how early in Longhorn’s life deployments must begin — Windows XP users will have more time. Microsoft will extend support for Windows 2000 (0.7 probability), but unless Microsoft supports it for at least three years after Longhorn ships (0.4 probability), you will have to choose to deploy Longhorn before you’re ready or use Windows XP as a stopgap. Microsoft should provide such guidance now, but because it wants enterprises to buy Windows XP, it will likely wait until closer to the end of the support dates before making an announcement. Gartner believes the uncertainty about Microsoft’s support for Windows 2000 until Longhorn can be comfortably deployed, coupled with security improvements that are slated for Windows XP with Service Pack 2, makes the third option — adopting Windows XP on new PCs — the best choice for most enterprises.

Should I standardize on a Windows version?

Adopting new client OSs through PC hardware attrition will result in a mixed environment, with some users running Windows 2000 and others running Windows XP. The ultimate goal of PC deployment is client OS homogeneity. However, homogeneity is expensive to achieve and fleeting. It often results in large inventories of defunct desktop products as enterprises stretch the life cycle, attempt to justify large migration projects and delay adopting technology. Most larger enterprises should have a strategy of managed diversity and OS replacement through hardware attrition.

Smaller enterprises (fewer than 1,000 users) may not have the economy of scale to manage a diverse environment. Thus, most should remain homogeneous. Also, enterprises that use large amounts of internally developed applications should use a homogeneous client OS because ISVs are responsible for ensuring that their applications run on multiple OSs, and most ISVs support their applications on more than one version of Windows.

If you develop applications in-house, your developers must ensure that the applications work on all of the OSs that you run. The cost of this testing and support can be sizable. “The Cost of Client-OS Diversity” (DF-15-3047) and “Client-OS Diversity: Application Development Costs” (DF-19-4424) present a model for understanding the cost of developing applications for multiple OSs and explore additional support costs. Use this model to explore the cost of supporting a diverse environment and compare it with the cost of maintaining homogeneity.

Should I move to Linux right after my Microsoft Enterprise Agreement ends?

The simple answer is “no.” The vast majority of Microsoft Enterprise Agreements (EAs) include perpetual rights to use the products in the agreement. They also include the right to run the latest versions of the products covered. With Microsoft’s seven-year support cycle for commercial software, most products will be viable for use long after the EA expires. Therefore, the decision to renew an EA or switch to Linux should be separate. If you don’t believe the EA offers sufficient value, don’t renew it. You should continue to use the products and even deploy the latest versions for which you have the rights for as long as possible.

The decision to switch to Linux should be made several years after your EA expires. Depending on the size of the enterprise, it will take a year to prepare a major OS or office productivity application migration, and two years to roll it out. Therefore, three years before the expected end of viability of your OS, assess what your future environment will be and begin your migration. This will give Linux, StarOffice and OpenOffice.org more time to mature. Microsoft’s Longhorn products will also be at a better point for evaluation.

If you’re serious about Linux on the desktop, a year is not enough time to move many applications off Windows. Applications that are being rewritten now should favor architectures that will enable flexibility in client OS choice. However, if many applications are being rewritten solely to move to Linux on the desktop, this migration will not make sense from a financial perspective.

Some enterprises say they want to move off Windows and Microsoft Office because they are unhappy with Microsoft’s licensing policies and want to exact revenge. For those enterprises, Gartner believes the best revenge is to continue to use Microsoft’s products, but stop paying for them by not renewing their EAs. (Of course, most enterprises will continue their support contracts even if they don’t renew their EAs, but at least they won’t pay licensing fees.) There is no return on investment in a migration done for spite, even if there is satisfaction in revenge (see “Linux on the Desktop: There Is No ROI in Spite,” DF-18-9810).

Will Linux on the desktop save me money? Where should I consider it?

Hype about Linux on the desktop is increasing. However, several qualities attributed to Linux will likely prove to be myths, such as:

I won’t feel forced to upgrade older versions of Linux. Eventually, support will be difficult to find because software developers won’t likely support very old versions of Linux.
With Linux, I can use StarOffice or OpenOffice.org instead of paying for Microsoft Office. StarOffice and OpenOffice.org also run on Windows.
Linux is free. Supported versions will cost money.
Other myths are discussed in “Myths of Linux on the Desktop” (DF-19-6831).

There is more to total cost of ownership (TCO) than hardware and software costs. You may find that the savings over the life of a PC from running Linux instead of Windows are less than you thought they’d be, are a small percentage of the amount that you spend, and are outweighed by migration and support costs. Gartner’s TCO research indicates that the acquisition cost of a PC with Linux may be only $80 less compared to a PC with Windows Professional (see “Linux on the Desktop: The Whole Story,” AV-20-6574). This is a relatively small amount over the three- or four-year life of the PC. Furthermore, supporting applications with less-than-perfect compatibility with incumbent software (for example, StarOffice vs. Microsoft Office) can increase support costs and end-user operations to the point where a Linux PC could cost more than a Windows PC.

Although lockdown and manageability could help reduce the cost of Linux, the same could be said about locking down and managing a Windows desktop. Also, in enterprises where political or cultural issues have prevented Windows lockdown, Linux lockdown may also be unattainable, unless it can be used to enact a culture change.

Migration costs also must be considered. For knowledge workers who run a wide variety of applications, the cost will be immense to migrate applications to run under a generic browser or native Linux. Instead, identify users who don’t run many applications as good candidates for Linux on the desktop. For example, data entry workers or limited-use structured-task workers could run Linux with less change and lower migration costs. Fight the temptation to make Linux desktops an “all or none” decision, and select the right users to migrate.

Written by Edward Younker, Research Products

Analytical source: Michael Silver, Gartner Research