Analyst firms have been releasing total cost of ownership calculations for leading operating systems, with Linux and Windows neck-in-neck for the low end, and Unix numbers somewhere up in the stratosphere. The question is, will standard TCO measurements still be relevant as we move into an impending era of utility, on-demand, and grid computing?
To corporate IT managers, Linux represents an interesting hybrid between Unix and Windows environments. Linux runs on commodity priced hardware (Intel, AMD-based boxes) just as Windows, but often requires more sophisticated expertise from the Unix world to put all the pieces together. Things get murky, then, when considering total cost of ownership costs, and to some extent, return on investment for these systems.
To get any new project through IT these days, one must pony up either TCO or ROI numbers. Since it's almost impossible to calculate potential ROI for an operating system implementation, analyst firms have been concentrating on tracking TCO numbers for Linux versus Unix versus Windows. What they're finding is Windows and Linux cost the same in hardware, with support costs tipping the equation one way or the other. However, analysts do not agree on whether Windows or Linux support costs more over the long run. Commercial Unix is another story, which analysts say just plain costs more in both hardware and support. Linux - and some flavors of Unix - win hands-down in upfront software acquisition costs, but this typically only represents only 10% of TCO over a three to five-year lifecycle.
A recent study by IDC finds that the TCO for a Unix system running on a RISC processor was up to six times higher than that of Linux running on Intel. In this study, funded by IBM and Red Hat Software, TCO is based on up-front acquisition and support costs. In a "collaborative" environment, the cost for Unix was $1,407 per user, compared to $256 for Linux. In an Internet or Web-based setting, the difference narrowed to $685 for Unix versus $377 or Linux. IDC cited hardware costs, accounting for the fact that older Intel-based processors could be recycled as Linux servers. Administrative costs also were higher for Unix, due to greater complexity.
Linux did not compare favorably against Windows in a separate IDC study. The study, underwritten by Microsoft, concludes that Windows is cheaper for functions such as networking, file/print and security, but Linux is cheaper for Web serving. At the high end, IDC estimates that running a file server will cost a typical Windows site $99,048 over a five-year period, versus $114,381 for Linux user. A Web server costs only slightly less to run on Linux - $30,000, versus $32,000 for Windows.
Another analyst firm, Robert Frances Group (RFG), brings Sun Solaris into the mix, and draws a different set of conclusions about Linux and Windows. In estimating the cost of a typical server environment, RFG says that the cost of hardware for both Linux and Windows server software averages about $39,000 over a three-year period. By comparison, a commercial Unix server box from Sun Microsystems will cost almost $400,000. Factoring in maintenance and support, the long-term costs over three years equate to $74,500 for Linux, $190,700 for Windows, and $561,000 for Solaris.
To account for some of the high Solaris costs, Chad Robinson, author of the RFG study, points out that "since we were considering long-term, mission-critical Web deployments, on Sun hardware these tended to be made on two to three redundant, vertically scaled systems. This did affect the salary totals because the server/admin number was affected."
Also, these studies looked at multi-user systems supporting relatively basic IT functions, such as Web serving and file serving. Unix proves itself in high-end applications over multiprocessor platforms. The TCO numbers could also be drastically changed by the fact that new releases of systems such as Solaris are backward compatible to much older hardware, which is not the case with Windows. As these TCO studies show, however, Linux is capable of eventually bringing Unix power to commodity-priced environments.
The real wild card in all the equations is the rise of utility, on-demand, and grid computing, or what Sun Chairman Scott McNealy often calls his concept of a "Big Freakin' Webtone Switch." Essentially, end-users will not have to purchase systems outright, but access resources across the network, be they off of a Solaris-based cluster, a Linux server farm, or Windows. While the concept still appears to be in its early stages, we're seeing early implementations and productizations of such capabilities from IBM, Oracle, and HP. Economies of scale may take hold, and the TCO playing field between Linux, Unix, and Windows may begin to level off. That's enough to make any corporate bean counter sit up and take notice.
Got something to add? Send me email.
More Articles by Joe McKendrick © 2011-05-02 Joe McKendrick
There are two ways to write error-free programs; only the third one works. (Alan Perlis)