Tuesday, December 17, 2002

Microsoft-sponsored study on Win2K vs Linux is NOT all good news for Microsoft

Earlier this month, the press gave wide coverage to an IDC white paper, sponsored by Microsoft, which found that (surprise!) Microsoft Windows 2000 (Win2K) is generally less expensive on the basis of total cost of ownership (TCO) than Linux.

The study of 104 U.S. companies analyzed spending trends between Linux and Win2K server environments. IDC found that Win2K is cheaper in four categories of use: networking, file serving, print serving, and security services. But the study did find that Linux has a cost advantage when used for Web hosting.

Microsoft's sponsorship of this white paper may be part of a larger PR campaign, since it comes on the heels of a recently leaked internal memo that indicates Microsoft is changing its tactics from denouncing open source software, such as Linux, to emphasizing the benefits of its own products. In light of Microsoft’s PR strategy, it’s tempting to discount the white paper as just an ad campaign for Win2K. Unfortunately, most reporters aided in that effort by simply repeating the conclusions of the study, as released by Microsoft, without examining the full report itself. So, I decided to get a copy of the full white paper.

Not that Microsoft or IDC makes it easy. Neither party is publishing the full report on its Web site. But eventually I was able to get a copy from Microsoft’s PR firm. [Update, 6/18/2003: I have been informed that the full report is available here on Microsoft's Web site.] After reading the lengthy document, it became clear why Microsoft’s strategy is to focus on the report’s conclusions — because in the body of the white paper, IDC makes at least three observations that are not favorable to Microsoft.
  1. Microsoft's advantage in administrative staffing costs is temporary. IDC observes, rightly, that the greatest cost category is people. The authors conclude that Microsoft has the advantage because Win2K is a "mature computing platform," for which there are more IT professionals and system management tools available, leading to lower people-costs. But the authors balance this conclusion with a note that Linux is an "emerging platform" and as Linux gains wider acceptance this gap will close:
    …as Linux matures and more packaged software becomes available in the Linux server market, IT professionals will be come more skilled in the efficient installation, deployment, and maintenance of Linux server environments (p. 4)....It is reasonable to expect Linux to support a more mature computing environment over the next few years, gaining better ISV support for commercial applications and packaged database products. System management tools are emerging and can be expected to expand rapidly over the next year or two in capability and installed base. It is also reasonable to expect less customization and scripting to be required for Linux computing over time, as Linux tools mature and become easier to use, thus reducing the TCO for the server environment (p.19).
  2. Linux is more reliable. In its analysis of file server TCO, IDC points out that W2K users experience considerably more downtime than Linux users, which drives up the cost of downtime in the TCO calculation.
    Taking a closer look at the data, IDC can state that, despite the vast improvements of Windows 2000 over Windows NT, the downtime associated with Linux servers is considerably less—often well less than half the downtime that users experience with Windows 2000 (p. 13).
    IDC explains that this may be partially attributed to the fact that Win2K servers typically carry larger workloads and that the interaction of multiple applications may cause the increase in downtime. Of course, one could argue that the free cost of Linux software enables IT departments to run as many copies as desired, cost-effectively spreading the workload across a greater number of servers. Furthermore, the free nature of Linux encourages its deployment in fail-over clusters, which shields users from drops in availability, as IDC itself points out in its analysis of the cost of downtime (p. 6).

  3. Linux is already cheaper for Web hosting. This is the one point in favor of Linux that did reach the conclusions disseminated by the trade press, although most reporters seem to view it more as a consolation prize for Linux. Nevertheless, Web hosting is the one area where Linux already enjoys a market share advantage over Microsoft, with the open source Apache Web server currently used on about 65% of the active Web sites, to about 25% for Microsoft’s IIS, according to a survey conducted by Netcraft. The fact that Linux has a TCO advantage over Microsoft in the segment where it has the most market penetration does not bode well for Microsoft in segments where Linux holds a smaller but growing share of the market.

    In explaining why Linux is cheaper than Win2K for Web hosting, IDC postulates that the cost of other Microsoft components, such as Microsoft SQL Server and Active Directory, are probably being lumped in with the cost of IIS even though those components can be leveraged for use outside of Web hosting (p. 17). This is a plausible explanation, although it inadvertently makes the case for Linux, since when an organization makes the decision to deploy a Microsoft platform for one purpose it tends to adopt Microsoft in other areas because of the tight integration with "common components." IT decision makers know this, and this could be a significant part of the business case for why Linux/Apache is so attractive for Web hosting.
Reading between the lines of the IDC white paper, nearly all the TCO advantages of Win2K over Linux are because Win2K is more established, with more system administrators trained in it and more system management tools supporting it. In other words, Win2K costs less to use because more people use it. At best, this is a temporary advantage, because as IDC notes in several places, Linux has been gaining market share at the low end, or as IDC puts it, on "one or two processor systems that perform relatively simple tasks on the edge of networks" (p. 6).

IDC acknowledges that, at some point, Linux will narrow the gap in trained system administrators and system management tools. This is consistent with Linux being a disruptive technology, as described by Clayton Christensen of the Harvard Business School. According to Christensen, a disruptive technology is one that is intrinsically lower-cost and simpler, but currently inferior to that of the market leader. It first gains market share among the least demanding customers or applications, those who are over-served by the dominant technology. As the functionality and performance of the disruptive technology improves, it moves up-market, gaining market share among more demanding users, eventually displacing the dominant technology. Without saying so, IDC identifies Linux as a disruptive technology, in the same way that mini-computers displaced a segment of the mainframe market, and networks of personal computers running Microsoft operating systems displaced a large portion of the mini-computer market.

Interestingly, in promoting TCO as a positive benefit for its Win2K operating system, Microsoft is making the same argument against Linux that IBM likes to make in favor of its AS/400 (iSeries) platform against Microsoft. A study conducted in 1998 by IDC, showed that users of IBM’s AS/400 enjoyed a lower TCO compared to both Microsoft and Unix platforms, due mainly to IBM’s significantly lower staffing requirements. Now the same argument is being made on behalf of Microsoft over Linux. But, just as the AS/400’s lower TCO did not slow the advance of Microsoft, it is unlikely that Microsoft’s lower TCO, if true, will slow the advance of Linux.

Linux is likely to continue to gain ground as a mainstream alternative to Microsoft operating systems. IT decision makers should evaluate Linux and Microsoft on a case-by-case basis, considering not only TCO but the specific workload requirements, existing and future staffing needs, and the organization’s strategic direction for IT infrastructure.