Microsoft’s Get the Facts site, which I discussed in last week’s LinuxInsider column, “Getting the Facts About Windows and Linux,” makes the case that Windows is cheaper than Linux. The site includes a contribution from Meta Group dedicated to the proposition that the combination of Linux plus a database costs more than a Windows Server plus SQL Server.
As proof, they adduce an up-front cost of US$4,000 for Windows 2000 Advanced Server plus $5,000 for SQL Server in contrast to Red Hat Enterprise Linux at $3,200 for one-year premium support plus Oracle at $15,000 or DB2 at $7,500. That comparison is deceitful by both omission and commission, but it does raise an interesting question: What does Linux really cost?
For many people, the right answer is simply that Linux is genuinely free. If you know Linux and already have a CD set, or can borrow one, you can install as many copies as you like without paying a nickel for licensing. Strictly speaking, therefore, a comparison between the cost of the Linux license and the cost of a Microsoft server operating system license should have a zero on the Linux side and some number greater than $799, depending on hardware, usage and the number of clients allowed, on the Windows side.
From Zero to $100K
Suppose, for example, we consolidated 40 Windows servers split evenly between 20 limited-usage file-and-print servers at $799 each and 20 database or application servers at $3,999 each to perhaps 16 Linux servers. In this case, the up-front cost comparison for the OS component should show Microsoft’s licensing at $95,960 and Linux at zero.
If you don’t already have or can’t borrow a CD set, you can either get the complete seven-CD Debian set delivered for less than 30 bucks or you can buy CD blanks and make your own if you don’t mind spending the time and have the bandwidth for a 4.5-GB download. In that situation, the comparison mentioned above doesn’t come out as $95,960 to nothing, but the Microsoft cost still looks like a bit of a loser at more than 3,000 times the cost of Linux.
In most cases, however, companies don’t have either the IT skills or the trust relationships between IT and business management needed to pull off a Linux server consolidation without incurring significant additional costs. In the best cases, those costs will be primarily related to the time needed to bring in the missing skills, vitalize IT and build those trust relationships, but few companies know enough about Linux to realize it isn’t Windows — meaning that MCSE ideas about computing don’t apply.
Defenestration the Key
What’s needed is defenestration — throwing out the Windows mindset along with Microsoft’s licenses and software — but that’s not as simple as changing a boot CD or even migrating a whole raft of servers. What’s involved is fundamental change both in how IT operates and in what it does. For example, the consolidation mentioned earlier would be a small part of a planned IT evolution with applications change, increased user control, some new skills brought in, some existing staff let go and some staff retrained.
That’s where the organizational benefits to Linux come from — not from a one-time capital cost reduction, but from long-term IT cost control coupled with an increase in IT’s ability to contribute to organizational productivity.
Unfortunately, you don’t usually see companies taking on carefully planned transitions to new ways of thinking about computing. Instead, mounting user dissatisfaction with IT services gets expressed among senior managers as rising concern over IT costs, and the two groups work out an unstated compromise in which the same IT people who can’t make the Windows stuff they know work to the satisfaction of their users promise to explore the use of Linux as a way to cut costs paid to third parties.
When they do, the implementation usually becomes a trial by error, with the MCSEs in charge attempting to use Linux as a cheaper substitute for Windows without making any other changes in their own behaviors.
Cost Is Misplaced Knowledge
As a result, the new systems never perform quite as well as the old ones, and costs start to mount as the people making the technical decisions look at everything except their own certainties about computing for solutions. In this situation it’s common, for example, to see one-for-one Windows-to-Linux conversions, with the case for premium support on every production server bolstered with every failure. In the case of our 40-server company, this combination reduces up-front savings to around $30,000 while creating a situation in which the cost of support forms the basis for a return to Microsoft in the second and subsequent years.
The worst part about this is that the MCSEs involved are right: Linux can never be as good a Windows 2003 server as is Windows 2003. Linux isn’t a cheaper form of Windows; direct substitution without change in thinking fails initially from a technology perspective and fails in the longer term from a cost perspective.
So, what does Linux really cost for business? Thirty bucks for the CD set plus everything your MCSEs think they know about how computing works in business; it also costs everything your MCSEs know about Microsoft technologies, starting with things like rack-mount SMP simulation, domain management, switch-based networking, DHCP, security, network or workload balancing, and server clustering originally developed to paper over gaping holes in Microsoft’s products — and therefore inapplicable to Linux.
Paul Murphy, aLinuxInsider columnist, wrote and published The Unix Guide toDefenestration. Murphy is a 20-year veteran of the IT consultingindustry, specializing in Unix and Unix-related management issues.