% fortune -ae paul murphy

Linux versus Windows: another fine Microsoft TCO Analysis

Almost a month ago, on September 24th, fellow zdnet blogger Mary Jo Foley filed a report under the title: Microsoft: Windows and Linux offer same TCO in emerging markets, which she introduced as:

While Linux may be cheaper up front, but over time, Windows and Linux offer roughly the same total-cost-of-ownership to customers deploying large numbers of PCs in schools in emerging markets.

As she points out, this Microsoft conclusion comes from work done for Microsoft by a company called "Vital Wave Consulting", so I thought I'd read the whole thing (PDF).

By volume most of the report is made up of:

  1. many assertions that using computers in schools improves education and economic opportunity;

    For example (from the summary):

    Academic research and private-sector investment decisions indicate that computers in schools contribute to improved academic outcomes, boost a nation's economic competitiveness, and attract job-creating economic investments.

    Lots of people say this - and this report quotes a few of them - but it's all opinion. The actual evidence is at best inconclusive, and suggests that the relative effectiveness of dollars spent on teachers versus dollars spent on computing infrastructure for education depends mainly on local social and economic conditions.

  2. beautifully formatted, colorful, and highly detailed looking tables;

    Unfortunately these seem to have magical elements - Table A, on page six, for example, gives a TCO that is not the sum of the listed components - and prices "mainstream computers" at $750 each over five years, but lists "second hand" computers at $254 for the first three years ($84/yr) and their replacements at $267 ($133/yr) over the next two.

  3. a lot of text that seems to be about the numbers - for example:

    Electricity and connectivity are two other areas where dramatically higher local costs could have a large impact on TCO. While electricity usage is a relatively small portion of the 5-year TCO for the conditions assumed in this study, power considerations can be critical in developing-country scenarios where electricity prices and availability are volatile. A school with electricity costs of US$0.80 per kilowatt hour, which is ten times higher18 than this study?s assumed price of US$0.08 per kilowatt hour, would have electricity costs of nearly US$4,000 over a five-year period. This would put electricity costs into the top cost categories, being nearly equal to the initial purchase price of the 16 ultra low-cost computers (US$ 4,560). Similarly, in areas that rely on satellites to provide Internet connectivity, Internet access costs can total up to US$60,000 over five years, more than doubling the TCO in all scenarios.

  4. an executive summary laying out the premise that Linux and Windows cost about the same:

    • Support and training are recurrent costs that constitute two of the three largest costs in the total cost of ownership model. They are greater than hardware costs and much higher than software fees.

    • Ultra-low cost computers and Linux-based solutions are relatively equal in cost to traditional hardware and proprietary software solutions because they require higher labor and replacement costs over a five year period.

    • The total cost of ownership for different computer types and software platforms is relatively consistent. Critical success factors, such as IT ecosystems and platform stability and longevity therefore need to be considered in affordable computing decisions.

Read this carefully and I think you'll agree that there's lots of hand waving disguising conclusions drawn from one belief: that it costs more to keep a Linux computer working than a Windows computer.

Since that's the key to the report's ability to off-set the costs for the Microsoft licensing, you'd expect some detail - here it is:

IT Technician Daily Rates

Assume a daily rate for IT technicians of US$64 ($8 per hour) for Microsoft-trained IT professionals and an 8.5% higher rate, US$69.44 (US$8.68 per hour) for Linux-trained IT professionals. Assume that the daily rate for rural areas is 10% higher. Data on the costs of hiring Linux- and Microsoft-trained professionals indicate that in both developed country and developing-country contexts, Linux-trained professionals command higher salaries and are thus more expensive to retain.

Salary surveys conducted in the U.S., Britain, and Australia by payscale.com consistently show that Linux-certified professionals earn 10-20% more than their Microsoft counterparts. The Red Hat India website claims that Red Hat- certified professionals earn up to 30% more than Microsoft-certified professionals (RHCE:

Although evidence is more anecdotal, the salary gap in countries such as China, India, Pakistan and South Africa appears to be similar, if not higher, that that in developed countries. According to a headhunter at a leading Chinese recruitment firm, a senior developer at a major foreign software firm, and the head of a medium-sized Chinese software firm (approximately 50 staff), salaries for Linux IT professionals are from 30% (systems administrators) to 43% (software developers) higher than salaries for Microsoft IT professionals in the same categories. Linux-experts in South Africa estimate that the differential between Linux and Microsoft IT professionals is closer to 10-20%, with Linux professionals earning the premium salaries. A global staffing firm and a high-tech company with affiliates in India as well as a local IT services firm estimate that Linux IT professionals in India earn up to double their Microsoft Certified Engineer peers. A major staffing firm in Pakistan provided survey data that found that Linux IT professionals with three years of experience earn 12.5% more than Microsoft professionals with the same number of years experience.

IT technician salaries were calculated based on an analysis of salary surveys (IT salary benchmarks from 2006 in Asia, http://www.zdnetasia.com/techjobs/salarybenchmarks/), IT technician jobs posted on career websites in Latin America and data from IT experts in China, India, Pakistan and South Africa.

As I see it, this first assumes the conclusions it draws on relative salaries, and then rounds out the net cost argument by making two related assumptions: that Linux needs as much support as Windows, and that the extra money allegedly paid for the Linux geek doesn't buy anything in terms of productivity or system stability.

When I looked briefly at job ads around the U.S. recently what I found was the opposite: no significant salary premium for Unix workers combined with significant productivity premiums for employers hiring them.

I don't know of a credible study on this, but it seems most likely to me first that salary differentials follow local market conditions more than they do technology specializations, and second that people with significant Linux expertise are far more likely to have other expertise to contribute than Windows administrators.

So what's the bottom line on the review? simple: if you hire someone to prove that increasing upfront spending on school computers by about 30% without getting a productivity improvement is a good thing, they're going to have to make some pretty slick logical leaps somewhere - and I think this one got to its destined conclusions by using nicely formatted Excel tables and lots of verbiage to hide assumptions amounting to the required conclusions.


Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specializing in Unix and Unix-related management issues.