Figuring out the total cost of ownership (TCO) for Linux versus Windows for any given application is not a simple matter. The answer depends on how an IT organization is set up, its in-house skill set and existing architecture, and a whole plethora of other factors.
The biggest problem is, most IT shops don't do full-blown TCO analyses for most projects, and many never will.
That's the consensus of industry watchers commenting on a recent white paper by International Data Corp. (IDC), which stated that the five-year total cost of ownership of Linux was between 11% and 22% higher than that of Windows 2000. This was true for four out of five workloads analyzed, IDC said, including networking, print serving, file serving and security. In one area, Web serving, Linux came out 6% less expensive than Windows 2000.
Factors studied included up-front acquisition costs for hardware, the operating system and applications software, as well as the costs of unscheduled downtime. The biggest single element pushing Linux over the TCO edge was staffing -- Linux requires more technical expertise than does Windows, and the cost of personnel needed in the Linux environment is simply higher than that for Windows, the 25-page white paper said.
Although the IDC white paper was sponsored by Microsoft, IDC used the same TCO methodology it uses in all its other survey work, whether or not the work is paid for by a vendor, said Al Gillen, an IDC analyst and one of the white paper's primary authors. "We've had great success with our methodology," he said. "If we couldn't stand up and defend this study, then it's worthless."
In this case, IDC surveyed IT managers from 104 North American companies. The customer names were selected from a list provided by Network World, which, like IDC, is owned by International Data Group (IDG). Microsoft did not have any input into choosing the customers selected, according to Gillen. Salary information for Linux and Windows administrators and other personnel came from salary surveys in Computerworld, another IDG journal, as well as other publications, he explained.
All told, the white paper "represents a composite of the 104 companies we interviewed," Gillen said, "and it provides a representative set of numbers. Nobody's going to look at this and say, 'gee, we're going to save X amount for this particular workload.'"
Instead, "it's my hope," Gillen said, "that people will give some thought to the kinds of costs they need to be looking at. The study can serve as a wake-up call to see where the real costs of operation are for any environment."
Most IT organizations don't do any kind of in-depth financial analysis, except for with their most strategic or expensive projects. And, typically, Linux doesn't fit this description, because the operating system generally comes into an organization slowly, brought by the techies who are using it to solve specific problems. The IT brass might not even know it's being used.
And even when shops do look at finances for a given project, it's most often done in terms of return on investment (ROI) -- what the company expects to get back from the project -- as opposed to what the project will actually cost over the course of five years. Almost no one does TCO analyses except for vendors and consultancies.
Tony Scott, chief technology officer at General Motors in Detroit, explained. "We do ROI, but we don't do it for every single project. The cost of doing a formal analysis can be more than the cost of the project" itself.
Instead, GM uses ROI as one factor in a broad portfolio-management process, in which IT projects are treated as investments, much as an individual would treat his stock portfolio. And the bottom line at GM is just that -- the bottom line. "We'll pick a numerical factor, to grow market share or reduce inventory, or reduce product development time," Scott explained. "The real test for us [is] if we can take those theoretical savings and apply them to the bottom line" to produce greater profits for the company and its shareholders.
Indeed, this approach to ROI is one reason GM has not yet invested heavily in Linux. "We've been able to avoid the herd mentality," Scott said. "We've had a number of Linux pilots, but we've not seen the vendors coming to us with a compelling business case. That's not to say there's not one there, just that we haven't seen one materialize yet."
Ted Schadler, a principal analyst at Cambridge, Mass.-based Forrester Research, says that the real issue in most enterprises is Linux versus Unix. "The Linux versus Windows hoopla is out of proportion, in my humble opinion," Schadler said. "Overall, there will be no persistent TCO or feature-function difference between Java on Linux or the Windows stack. The vendors will ensure that."
The more interesting issue is that "the price-performance of Linux-on-Intel makes Unix-on-RISC a dead dog," he said.
Whatever you're comparing, the questions are more about the quality of the architectures, developer tools, vendors' partner "ecosystems" (as Schadler calls them), and business productivity. And the answers will vary among different customers, for different reasons.