Windows vs Linux: Limitations of TCO analysis for IT decisions

Stacey Quandt and Jon Erickson

The Microsoft-sponsored IDC white paper titled Windows 2000 Versus Linux in Enterprise Computing is an attempt to identify the cost items to support each operating system environment, but it is of limited value since it cannot help users rationalize major technology decisions.

IDC interviewed IT executives and managers at 100 companies in North America and asked them to quantify the operational costs of servers running Windows 2000 and Linux across five workloads – network infrastructure, file serving, print serving, web serving and security applications.

The study concludes, with the exception of web serving, that the costs of Windows 2000 are 11 percent to 22 percent less than Linux over a five-year period. IDC found that for web serving, Linux had a 6 percent cost advantage over Windows 2000.

Some of the findings, such as the fact that the initial acquisition cost of an operating system is only one aspect of an IT decision and that staffing cost is the greatest contributor to IT budgets, resonate with published Giga research. Yet, overall the study is incomplete since it fails to look beyond total cost of ownership (TCO) to measure the full economic benefit of the two solutions.

Rather than highlight the advantages of a TCO analysis, the study shows the disadvantages of basing an IT decision on TCO. This stems from the fact that TCO is not an appropriate measurement for most IT decisions. This is not a new revelation. By measuring only cost outlays over a specific time period, IDC fails to assess the full impact of benefits, flexibility and risk of an existing and alternative technology.

TCO can be beneficial in setting budgets, but it is not a useful tool to compare the changes to the business delivered through an existing operating system or an alternative approach. In particular, the deficiencies of this IDC TCO study are as follows:

First, IDC pointing out that Linux is not free: There are always costs in deploying and maintaining any operating system environment – including Linux. What is missing from the IDC study is information on the Linux kernel versions and specific Linux distributions in use by the companies interviewed.

This information would provide valuable data on the complexity and costs of supporting each workload. For example, the cost of acquiring a Linux distribution can range from $0 to $2,500. Linux can be downloaded from the Internet and is in this sense "free"; however, the actual purchase price of a Linux distribution merely covers the cost of CDs, documentation and shipping, not the actual software.

Hence, a study that attempts to compare Windows 2000 with Linux is akin to comparing Coca-Cola with fruit juices.

Second, the TCO metrics describe a five-year time frame. Giga believes that the time frame considered should be either two or three years. We do not recommend IT infrastructure initiatives be evaluated beyond this point because possible changes to technology and business models makes it difficult to estimate costs beyond this time period.

For example, the recent Microsoft Software Assurance Program has an impact on

the costs associated with a Windows system. It is impossible to say that within a specific time frame such licensing practices will not occur again.

A third issue is IDC’s conclusion that Windows 2000 is a mature operating system coupled with shrinking costs and that, conversely, Linux is immature and requires more work to configure, program and support. Who did IDC actually talk to – Windows or Linux users? This is an issue of aptitude. Linux systems can be easily configured using DHCP, LDAP, shell scripts and third-party products.

Mature computing platforms do not always have an advantage in cost measurements. The Windows operating system changes radically every few years. There are costs associated with change. For example, the use of Active Directory adds more complexity to a Windows 2000 operating environment, and there are costs associated with training system administrators.

Fourth, the IDC study fails to look at the value of flexibility of either solution.

Not being able to measure the value of flexibility is an inherent shortcoming in TCO that is especially important for infrastructure decision-makers. In comparative analysis between Microsoft and Linux, the flexibility value of both must be quantified. For example, in the case of Microsoft, the ability to inter-operate with other solutions is an important consideration that must be quantified. For Linux, the multiplicity of various Linux distributions providers and the value of hardware portability is an important consideration that should be part of the economic justification.

Before embarking on any large-scale technology investment, organizations must be

able to quantify the strategic value of flexibility. Measurement on the basis of cost and cost savings limits the decision to a tactical decision where a strategic solution may be necessary.

Fifth, IDC fails to show how different risk factors specifically impact the financial return of the two solutions.

Organizations need to look at how risky it is to stay with their existing infrastructure and should be able to quantify the risk of migration to a new operating system environment. In addition, any risk analysis must take into account the likelihood that the benefits that are created are measured and demonstrated to show a direct impact to the bottom line.

Specific risk factors such as vendor support, training, software licensing and downstream legal issues are important risk factors that must be considered.

Overall, TCO is only useful when evaluating products and methodologies related to systems and architectures to which your organization is completely committed. A TCO analysis should not be used to justify major technology or architecture directions, to set priorities for major products, or to compare products in different categories.

A forward-looking IT manager should assess the benefits, costs, flexibility and risks that Windows 2000 and Linux will provide in driving the organization to its long-term goals.

Giga recommends organizations use a model that accounts for the risk of changing or not changing the current architecture. It should include mapping the IT direction with the corporate direction. By merging the total current cost with the flexibility and benefits and tempering these with the risk, IT can achieve important strategic decisions.

This will not only increase the organization’s worth but also transform IT from an overhead cost into a valuable asset.