It's one of the perennial questions facing the open source movement: Is Linux ready for the corporate desktop? Ready or not, Linux is coming.
Industry research company IDC predicts that enough companies will see the benefits of a Linux desktop to increase paid shipments of the operating system from 3.4 million clients worldwide in 2002 to more than 10 million by 2007, giving Linux a small but respectable 6 percent of the desktop market.
"Linux captured the No. 2 spot as desktop operating system in 2003," says IDC analyst Dan Kusnetzky. "IDC sees Linux maintaining that No. 2 position and growing ever so slightly -- but not becoming a dominant force or even a major force."
Regardless of whether Linux will be a real threat to Windows on the desktop, the expected growth was enough to prompt Linux vendor Red Hat Inc. to introduce an enterprise-focused Red Hat Desktop product in May. Despite being the leading commercial Linux vendor, Red Hat actually lags behind Sun Microsystems' Linux desktop sales to enterprises. Last fall, Sun began shipping its Java Desktop System, which is based on the SuSE Linux distribution. And SuSE itself has shipped a product for corporate users since March 2003, called the SuSE Linux Desktop. Since being acquired by Novell earlier this year, SuSE has been readying the Novell Linux Desktop, which will incorporate software from Novell's 2003 acquisition of Ximian.
"We're listening to customers, and we're listening to the market," says Novell CTO Alan Nugent. "And the market keeps saying, 'Why do I have to spend all of this money on licenses for XYZ software when these 40,000 desktop machines are sitting here just running a mail client, a Web browser, and a couple of Citrix-served apps?' "
Good question. Unfortunately, it's still extremely difficult -- if not impossible -- to get solid data on Linux's TCO on the corporate desktop. And although the Linux desktop has strong appeal for some customers, it isn't for everyone.
In January of this year, Eddie Bleasdale, a systems integrator based in Surrey, England, formed the Incubator Club, an informal group of IT managers who meet regularly to discuss their experiences with Linux and open source migrations. According to Bleasdale, current members of the club are engaged in pilot projects and even some large-scale Linux migrations, which could collectively total as many as 200,000 desktops.
According to Bleasdale, if IT managers take a long-term approach, the move to Linux desktops has enough benefits to make the switch worthwhile -- from lower software licensing costs, to a slower hardware upgrade cycle, to increased efficiency owing to more secure, scalable, and reliable systems. "Your objective should be that, over a period of years, you should be able to reduce your costs of computing by 50 percent," he says.
One major outstanding issue for any company looking to migrate to the Linux desktop, however, is application support. With command of more than 90 percent of the desktop market, the Windows operating system is deeply entrenched. When one considers the legacy applications and Microsoft dependencies common to most enterprises, a move to Linux can quickly become a complicated proposition.
Take the Allied Irish Bank, which recently began the process of migrating 8,000 of its branch-office clients from Windows 3.1 to Sun's Linux-based Java Desktop System. Linux was appropriate for the bank's branch offices, which have a "definable functionality," says Michael Bowler, the bank's IT architecture manager. But with approximately 500 Windows-based applications in AIB's corporate offices, the bank has no plans to deploy Linux companywide.
"Corporate head-office environments tend to have a large, disparate application set. I'm not sure you could define the functionality," Bowler says. "It's a big elephant to chew."
Outside of small-business and home-office environments, this is a common story for Linux on the desktop. Larger enterprises are adopting tentatively but only in places where administrators can use it for a limited set of applications and where it offers a definite and compelling benefit, in call centers, technical workstations, or on point-of-sale terminals, for example.
For AIB, those clear and compelling benefits came from centralizing data that had been distributed throughout the branch offices onto a more flexible back-end application architecture (see "Investing in the Linux Client," page 49). "Our view was that centralizing all of the data was the best way forward from a regulatory and compliance perspective," Bowler says.
Whereas complex, custom applications in particular may be difficult to migrate to Linux, more mainstream desktop productivity applications pose less of a challenge. The set of alternatives to Windows applications is growing, including Sun's StarOffice office productivity suite, the Mozilla browser, Novell's Evolution groupware client, OS virtualization software such as VMWare Workstation, and CodeWeavers' CrossOver Office, which allows users to run Windows applications directly on Linux. With such alternatives available, users can experience a Linux desktop that closely resembles the Windows they are used to.
Compatibility software such as CrossOver Office can be enough to push some companies over to Linux. Digital animation studios Pixar, DreamWorks SKG, and Disney have moved a combined total of 2,400 of their technical workstations to Linux. They use the CodeWeavers product to run software that the studios cannot port to Linux themselves, such as Adobe Photoshop, according to CodeWeavers CEO Jeremy White.
"All their animators run Linux," White says. "They used to have a Windows computer and a Linux computer, and now they just have a Linux computer."
CrossOver Office isn't the only solution for delivering Windows applications to Linux users. Citrix Systems Inc.'s MetaFrame Access Suite and Tarantella Inc.'s Secure Global Desktop can serve applications-hosted Windows machines to Linux clients. But all of these products require some tuning, or at least extra software licenses, to achieve parity with Windows. Even CodeWeavers' White agrees that Windows compatibility, in itself, still leaves users without a compelling reason to switch to Linux.
"Even if Linux were a drop-dead, perfectly compatible, just-as-good replacement for Windows, you wouldn't find a lot of people switching," White says. "You have to not just be good enough; you have to be compellingly different to take over the world."
Flipping the switch
For some, Linux is already more than good enough. Already in the process of moving the city's servers to Linux, the city of Bergen, Norway, expects to begin moving desktops in its 100 schools to the free operating system next year. For the city's head of technology, Ole-Bjorn Tuftedal, Linux on the desktop will mean less frequent hardware and software upgrades, a wider choice of software, and improved virus protection.
"There are many different costs that may not be apparent right away, but if you take them into consideration over a longer period, then they may pay off," Tuftedal says.
The city has already undergone two comprehensive user tests of Linux desktop systems. Users who were familiar with Windows were able to navigate through Linux's KDE (K Desktop Environment) and Gnome desktop interfaces. Linux proved to be a more secure architecture, as well. "Linux was easier to lock down, and the built-in security was stronger," Tuftedal says.
Tuftedal found that centralized system management under Linux had a number of advantages, including a good selection of administration tools. Patch management, for example, was easier and more reliable under SuSE Linux's YaST (Yet Another Setup Tool) online update software than with Microsoft's Systems Management Server 2.0.
"It's easier to set up all the clients to automatically download patches from a centralized server after the different patches have been vetted by us," Tuftedal says. But he's quick to note that the transition required some technical savvy. "Here the tools are used by skilled engineers where it pays to train them, if necessary, and the 'depth' of the tools is more important than a slick user interface."
It may take more than skilled administrators to support a large desktop Linux deployment, however. In some cases, hardware support can be an issue. Historically, support for the latest hardware under Linux has been spotty -- but that's changing.
Some improvements derive from advancements in the Linux kernel itself. The forthcoming Novell Linux
Desktop is expected to be based on the Linux 2.6 kernel, which has greatly improved support for plug-and-play devices when compared with the 2.4 kernel, according to Novell's Nugent, who says the difference is analogous to the difference between Windows NT and Windows XP. "(The 2.4 kernel) is a lot like NT in terms of desktop capabilities vis-a-vis plug-and-play," he says.
Improvements in the 2.6 kernel aside, most IT managers still agree that Linux lags behind Windows in hardware support -- something that is particularly noticeable in esoteric areas such as laptop device drivers. But according to Bergen's Tuftedal, Linux now has enough support from hardware manufacturers to be viable.
"Hardware support was more of a problem some years ago," Tuftedal says. "Even though Linux isn't very large on the desktop, its market share is probably twice what Apple's is, so it does make a difference for hardware manufacturers."
One area where Linux has a distinct advantage over Windows is in support for older hardware.
Robert Duncan III, a technologist at Bacone College, is using Linux to extend the life span of 15 of the college's older desktop systems. Using a combination of Linux, a memory upgrade, and new CD-ROMs on 133MHz Pentium systems, his team has turned what would otherwise be obsolete hardware into serviceable Internet kiosks for the college's students.
"Those machines are ancient in the Windows world and would wind up in some landfill if it weren't for Linux," Duncan says. "Our users don't usually even notice it's a Pentium 133. With KDE, a common desktop manager, it looks like high-end software without needing high-end specs to match."
When asked to name the main barrier preventing more widespread Linux use on the desktop, Bergen's Tuftedal sounds a familiar refrain. "The real problems are with your important applications," he says.
Although Bergen has spent the last few years trying to reduce the number of dependencies on the applications it purchases, the city still has a number of legacy applications that will not run on Linux. "We either will have to incur the cost of porting them, incur the costs of switching to alternative software, or we will run them on Windows terminal servers," Tuftedal explains.
User experience and expectations are also a barrier, says Bacone's Duncan. "Users are not as savvy as they should be, and they have this perceived value that open source software can't be worth anything because it's free."
Ultimately, says the Incubator Club's Bleasdale, a move to Linux is about one thing: freedom of choice. But he is quick to add that IT managers do not have to jump directly into Linux in order to make the first step.
"The thing you don't want to do is look at all of your applications that run on Microsoft and say, 'Gee, why can't we run all of them,' and then throw in the towel," Bleasdale says.
A more reasonable first step, according to Bleasdale, would be to examine the possibility of moving to the Mozilla Web browser. That, in turn, will expose any dependencies the company's Web-based applications may have on Internet Explorer and can eliminate the security risks posed by IE.
"Microsoft has put itself into a real corner by integrating IE into Windows," Bleasdale explains. "Microsoft is in a really very difficult position with the problems of Internet Explorer. In fact, I think their only way out is to buy another company with a Web browser."
One by one, administrators can begin to migrate existing Windows applications to Linux. The open source Samba file-and-print server software, for example, can be used to create a cross-platform networking environment, whereas the StarOffice productivity suite can provide a free alternative to Microsoft Office. The ultimate goals, Bleasdale says, are to reduce application dependencies and costs.
"(Migrating to Linux) is not something that can be done overnight," Bleasdale notes. "People are going to have to go out and justify going through what will be a difficult process."
Investing in the Linux client
User experience, application availability, ease of maintenance, and stability are often chief concerns when deciding whether to deploy Linux on the desktop. For the Allied Irish Banks, however, the decision to move the 8,000 clients in the company's U.K. retail banking network from Windows to Linux had more to do with the server than anything else.
That's because AIB's decision to upgrade its aging Windows 3.1 workstations with new systems running Sun Microsystems' Java Desktop System ultimately sprang from a decision to centralize all of the bank's data through Web-based applications running on a J2EE framework. "We are predominantly using (Linux) as a platform to deliver the Mozilla browser," says Michael Bowler, the bank's IT architecture manager. "The client operating system doesn't really matter from the perspective of delivering line-of-business functionality."
The rollout, slated for 2006, is now in the testing phase. AIB developers have exposed the bank's core financial applications, which run on a mainframe, as Web services. They are also developing systems management, document management, Internet, and directory services -- all of which are being written to deliver an OS-agnostic application platform that can be used with Windows or Linux clients interchangeably.
This architecture will give the bank centralized control of data, but Bowler believes it will also deliver improved security, stability, and manageability for the client systems and their applications. "We have lots of flexibility and lots of choice, and we're very much in control of our environment," he explains.
Although the choice of client system may not matter from a technical perspective, going with Linux clients will save the bank money on licensing costs and will help ensure that its developers "think in more vendor-agnostic terms," Bowler says. It also gives AIB access to a large variety of open source components. AIB engineers were already able to use open source software to create a custom-built software upgrade system.
As for user experience on the Linux clients, although it may not match the glitz of the latest Mac OS X or Windows XP desktops, Bowler isn't concerned. "We do not perceive there to be a significant risk there," he says. "We're talking about users coming from a Windows 3.1 base."
In search of the bottom line
Linux on the desktop might mean freedom from software-licensing costs for some IT departments. But when it comes to evaluating desktop Linux's TCO, it's the human cost that is most important. According to industry research company IDC, the staff-related costs of system administration, support, development, and training typically amount to between 50 and 70 percent of the costs of the average enterprise application, when measured over a five-year period. By comparison, software costs for both client and server software total only 8 percent to 10 percent of TCO, IDC reports.
The heavy impact of staffing costs may be an impediment to client-side Linux adoption, says Dan Kusnetzky, vice president of system software at IDC. "There's a concern, whether it's true or not, that injecting Linux into the environment may increase staff-related costs," he says.
According to Kusnetzky, even savings of $50,000 to $100,000 per year on software or hardware can be nullified if one new employee must be hired to administer the new software. "Anything in the organization that increases staff costs quite often submerges the savings on an individual item, even if the item was free," he says. "That's why companies tend to stay with the status quo."
A June 2003 TCO study by Gartner found that, although the move to Linux on client desktops could save enterprises $154 per user per year on hardware and software acquisition costs, these savings would often be offset by the cost of migration applications from Windows to Linux. "Because all Windows applications must be replaced or rewritten," Gartner found, "migrating desktops to Linux only makes sense in a very narrow, limited range of situations."
Time spent learning the new Linux environment and dealing with Windows compatibility issues could also reduce any cost savings, Gartner said. It found, however, that Linux had a more favorable TCO when compared to unsupported versions of Windows, such as Windows 95.
The report recommended a Linux migration only for a select few tasks, such as data entry or platform automation, and warned, "Migrations that are made mainly for political reasons or to spite Microsoft can be costly."
Some who have made the switch to Linux on the desktop say they believe that its stableness and remote manageability have actually saved on administration costs, but reliable data on the TCO of the Linux desktop is hard to find. Companies looking to make the switch will need to keep a close eye on staffing costs.