Yankee: Linux Will Grow, But Windows and Unix Will Persist

by Timothy Prickett Morgan

According to a report comparing the total cost of ownership of Linux with Unix and Windows by consultancy Yankee Group, Linux on the server is going to experience double-digit growth in the coming years, but that does not necessarily mean that Windows and Unix are going to fade away. As is the case in much of the IT industry, Yankee Group's study of IT buyers seems to indicate that platform inertia will keep Unix and Windows in dominant use for years to come.

Yankee Group and Windows system software supplier Sunbelt Software surveyed 1,150 IT administrators and C-level executives worldwide in early 2004 to get a sense of what they thought about Linux in relation to the Unix and Windows platforms that most of the world still uses to do data processing. (The two organizations did not seem to survey customers who exclusively used proprietary mainframe or midrange servers, which could somewhere around half a million unique users worldwide and represent a big portion of IT spending every year.) The two-part report issued by Yankee Group, appropriately called the "Linux, Unix, and Windows TCO Comparison," was based on a survey of 850 IT managers and C-level executives who answered 23 questions about their Linux, Unix, and Windows plans; companies of all industries, sizes, and geographies were invited to participate. In a follow-up survey that targeted larger companies (those with 5,000 or more end users), Yankee Group was able to get 300 responses. Yankee Group then did 24 in-depth interviews with IT managers in North America and Europe to get some anecdotal background information to flesh out its survey information. About a third of the survey respondents had Linux installed, either as a Web server platform or for specialized departmental servers.

While the information that Yankee Group has gleaned from IT shops is no doubt useful, the first page of both part one and part two of the TCO report appears to have a math error that skews the interpretation of the results. This first page has a pie chart that mixes and matches data from Unix and Windows shops as they discuss whether they will ignore Linux and keep their Unix and Windows, deploy Linux along side of Unix and Windows, and replace Unix or Windows with Linux. The pie chart shows percentage of respondents across these two platforms, not percentage of respondents by platform, and this leads to some surprising conclusions. For instance, across all of the more than 1,150 shops surveyed, 4 percent of shops said that they had no plans for changing the Unix installations, while another 3 percent of shops said they planned to add Linux servers but not replace Unix machines and yet another 4 percent planned to migrate their Unix boxes to Linux. The report mentioned a number of times that only 4 percent of Unix shops planned to migrate to Linux. That cannot be mathematically true. Without knowing the precise percentages for rounding, there appeared to be around 127 Unix shops in the surveys of 1,150 users. Given that Unix has been the dominant server platform for years out there in server land, this Unix data is pretty thin. But even if you accept that the data is accurate, a very large percentage of Unix shops appear to be thinking about moving, based on Yankee Group's data--if you do the math, it looks like something more like 36 percent of Unix shops are contemplating a wholehearted move to Linux, another 36 percent of Unix shops are staying put, and the remaining 28 percent are mixing Unix and Linux.

Similarly, the pie chart also shows that 33 percent of those companies surveyed said that they had no plans to change their Windows servers, with 25 percent saying they will migrate a portion of their Windows boxes to Linux, 21 percent saying that they will add Linux servers but not replace their Windows servers, and 11 percent planning a total jump from Windows to Linux. Again, these statistics are more relevant when recast not as percentages of the survey group, but as percentages of the group that had Windows boxes. Doing the math again, it looks like about 37 percent of Windows shops have no plans to change their Windows servers, with about 28 percent migrating some Windows boxes to Linux, 23 percent adding Linux alongside Windows boxes, and 12 percent planning a migration from Windows to Linux. (It is interesting to note that Yankee Group and Sunbelt data as depicted in this pie chart did not appear to ask what percentage of companies wanted to move from Unix to Windows or from Windows to Unix. These two are both done, particularly the former.) This data presented by the Yankee Group report fits what many of us see going on out there, and it is merely its presentation that can give the misleading impression that only a minority of Unix shops are looking at Linux. The majority of Unix shops--based on the Yankee Group/Sunbelt data--seem to be putting in Linux or have it already.

The Yankee Group report, echoing similar findings from other consultancies and anecdotal evidence from server makers around the world, identifies that Linux is particularly useful in "green field" installations where a whole new application is being installed and is particularly attractive to Unix shops who are tired of paying high prices for RISC/Unix iron.

Not surprisingly, the Yankee Group study indicated that large companies with complex IT infrastructures are hesitant to jump to Linux for the sake of Linux--as they well should be. "Despite the hype, Linux is not superior to Unix and Windows Server 2003," writes Laura DiDio, Yankee Group's application infrastructure and software platforms senior analyst. "Linux is the technical equivalent of Unix and Windows Server 2003. In large enterprises, a significant Linux deployment or total switch from Windows to Linux would be three to four time more expensive and take three times as long to deploy as an upgrade from one version of Windows to newer Windows releases." This assessment was backed up by 90 percent of the 300 large organizations surveyed by Yankee Group and Sunbelt, who agreed that such a switch would be "prohibitively expensive, extremely complex and time-consuming, and would not provide any tangible business gains for the organization."

And while Yankee Group concedes that the lure of cheap X86 iron (and particularly now as 64-bit servers are available from Intel and AMD) makes the jump from RISC/Unix to Lintel or Lopteron servers more appealing, the relative immaturity of Linux in terms of systems management tools, workload managers, and application software support is an effective barrier to a wholesale jump from Unix to Linux for many companies.

Perhaps more striking is the fact that skilled Linux administrators are, thanks to the explosion in Linux enthusiasm in the market, in short supply and right now Linux administrators can command a 20 percent to 30 percent salary premium over Unix and Windows administrators. So, companies might save money moving to X86 iron from RISC/Unix iron, but they may end up paying for it in higher training costs and salaries in the short term. (Linux servers may take fewer administrators than Windows boxes, but the admins cost more. Catch 22.) Yankee Group projects that Linux administrator salaries will level off soon and will be equivalent to the salaries of Unix and Windows administrators (which are in large supply) in the next 24 to 36 months. And while the patching, security, and hacking menace currently makes supporting Windows costly compared to Linux these days, the Yankee Group report says that the improvements Microsoft is making in the Windows server platform will make it easier to patch and secure, and therefore less annoying and less costly to support. Moreover, as Linux is more prevalent in data centers, more people will try to hack it, and it will very likely become more troublesome to support. (Linux zealots will counter that Linux, due to its open source nature, is inherently less hackable, but C-level executives polled by Yankee Group and Sunbelt often had the contrary opinion that closed source software like Windows would inevitably be more secure.)

The study also tried to get a sense of the reliability of the Linux, Unix, and Windows platforms as they are used, not in some theoretical lab setting. The respondents to the surveys behind the Yankee Group report tended to use Linux in Web infrastructure capacities, and 80 percent of the respondents said that their Unix or Windows workloads were double or triple that of their Linux machines. So an fair comparison is tough. However, perception about the stability of a platform on one job can often lead to its use on others--which is how Unix got to be the powerhouse of the server market in the 1990s and how Windows will very likely become the powerhouse of the 2000s. Some 76 percent of respondents said Linux and Unix had similar stability and reliability (as gauged by the number of reboots), and 70 percent of those responding to the survey said they believed Windows Server 2003 was equal to Linux in this regard. Windows administrators griped about how bad dealing with patch management and security issues were with that platform, and this, more than anything else, seems to have hurt comparisons between Windows and Linux.

The crux of the matter was brought home in part two of the study: Linux is a great bargaining chip if you are either a Windows or Unix shop looking to cut costs. In the study, Yankee Group cited the MIS manager of a midrange retail company. "We have no intention of switching to Linux, but we do find it to be a useful stone to throw at Microsoft."

There are other issues that will hold Linux back in the data center, according to the Yankee Group study. Red Hat has only certified 750 applications on its server platform, while both Windows and the various implementations of Unix all have thousands of applications available. Linux may also have lower licensing fees than Windows or Unix, but there is limited warranty and indemnification from possible legal issues pertaining to copyright or intellectual property. The latter presents either an actual hidden TCO cost (if you buy indemnification insurance today) or a future one (if you get sued as a Linux user). Even the much-discussed benefit of being able to tweak the Linux kernel and its applications to suit a company's needs can cut both ways. Such mucking about in the code can breach indemnification offerings and warranties, and it can also make it very difficult to patch the software. Making use of the freedom of open source software has its price as well, it seems.