Marking what could have been a summer-long hiatus in its "Get the Facts" campaign, Microsoft is re-igniting the flames on the argument over whether enterprises spend less to manage Windows systems than Linux systems.

This morning, the company touted a study it commissioned from independent analyst Mercer Management Consulting, which made the case that companies that implement migration programs away from UNIX systems based on the need to adopt new applications -- what Mercer calls "transformational migrations" -- now tend to choose Windows over Linux.

One of the paper's key claims is that businesses that choose Linux over Windows for their UNIX migrations tend to be less rigorous in how they ascertain potential TCO savings.

"Many IT executives at companies that migrated from Unix to Linux admitted, when interviewed for this study," the paper states, "that the decisions to move from Unix to Linux (rather than Windows) often are based on intuitive expectations of savings and preconceived notions about the ease of migrating specific workloads to Linux."

The paper quotes one IT manager representing the minority viewpoint as saying, "You don't have to do an ROI analysis to understand it makes sense to switch to a platform [Linux] that is a factor of 3 or 4 cheaper." Microsoft commissioned Mercer to perform the study.

In a press release/interview with Mercer consultant John Wenstrup, Microsoft quoted him as saying the cost of the operating system license is not the driving factor in TCO today.

"When some companies are considering migration," said Wenstrup, "they assume that the costs of the hardware and software platforms are the overriding cost elements. In fact, what we found is that most of the cost of migration -- around 80-85 percent -- is composed of the labor of executing the migration and the ongoing labor associated with managing the environment. The total cost of the entire hardware and software platform -- including maintenance costs -- tends to only be 15 to 20 percent of total migration costs."

The cost of the software platform itself, Wenstrup contends, is only about 2% of total TCO, or as much as 5% when ongoing software maintenance costs are factored in.

For Linux proponents' part, IBM continues to cite an August 2005 study it commissioned from independent analyst Robert Frances Group (PDF file here), stating overall costs of Linux ownership tends to cost businesses 40% less than Windows over a three-year period.

The cost of maintenance for Linux applications is still much higher, the Frances Group study contents, although that paper believes maintenance to be a lower percentage of overall TCO than hardware costs, but doesn't give a specific breakdown. The right hardware choice, the Frances paper implies, makes all the difference - and reminds readers that IBM commissioned the study.

The Mercer study contends hardware costs can constitute up to 11% of TCO, with the cost of migrating software and applications consuming up to 42% of costs, and ongoing administration costs about 28%.

The Frances study examined 20 clients who were using Windows, Linux and Solaris across their enterprise over a three-year period. All clients were running Java 2 Enterprise Edition (J2EE) on their servers, but it contends that the Windows and Solaris clients both tended to assign separate teams to the operating system and to the JVM, whereas the Linux-based clients pooled these support teams together. Here is where the Linux clients enjoyed the biggest cost savings, said the Frances study.

"The high cost of Linux is misleading, and does not indicate a difficulty in managing JVMs on Linux," states the Frances paper. Although it says the higher efficiency of Linux-based hardware does enable clients to purchase fewer servers to perform the same workload, it goes on to say the real cost savings is strategic: specifically, by employing fewer people to administer services, and creating fewer administrative departments to be delegated among a larger IT staff.

As Mercer's Wenstrup admitted to Microsoft, some of the clients in its study performed substantial TCO analysis, and did not actually come to the conclusion that either operating system was substantially less expensive than the other.

"Quite surprisingly to us, the vast majority of companies who performed even moderately rigorous TCO analysis made it clear that the risk-adjusted returns for migrations suggested Windows was often essentially a wash with Linux -- and sometimes even advantaged -- on total costs," said Wenstrup. He went on to quote one client who said, because Linux and Windows "provided about the same TCO," the company could make its choice based on applications and availability.

The fact that the IBM-commissioned study pointed to the low price of Linux-based hardware, and the Microsoft-commissioned study pointed to the advantages of being able to choose based on applications, should surprise very few.