Putting TCO Studies In Their Place - page 3
Analyzing the Analyst
For this report, commissioned by Microsoft, IDC surveyed North American medium and large companies only. ( A "medium" company, Kusnetzky defined, is one with 51-499 client systems. A "large" company has 500-plus systems.) The analysts interviewed over 100 IT executives from these companies, focusing on the workload areas of: network, file, Web, print, and security services.
After the data was gathered, Kusnetzky went on to explain, it was normalized per 100 users. A four percent increase in IT costs was factored in for each 100-user group, and projected out over five years.
While the study examined a specific Microsoft product, Windows 2000, it did not examine just one Linux distribution. Instead, an amalgam of Linux distros was examined. I asked Kusnetzky why this would be so, and he explained that to compare one company's product to another company's product would violate IDCs hard and fast "rules of engagement."
Under these rules of engagement, IDC will not compare one product to one product. Instead, a single product will only be compared to a range of products. This is done, Kusnetzky emphasized, to avoid any impropriety.
"Were very careful to avoid any appearance of bias," Kusnetzky said.
In the studys results, IDC found that in four of the five workload areas, Linux was "marginally more expensive than Windows 2000." Most of the added expense, Kusnetzky said, comes from staffing and training costs for Linux applications. In most IT organizations, Kusnetzky stated, hardware and software costs, where Linux traditionally shines, comprises less than 35% of the organizations IT costs. Staff costs, however, can be 50-70% of the total IT costs. If this IDC model is fact, then it would explain why Linuxs traditional strengths are outweighed by IDC's surveyed staffing cost figures.
Kusnetzky maintains that this is what IDC has maintained ever since it started surveying Linux and Open Source software--that the added costs in hiring and training Linux-proficient staff will offset the free nature of the software.
But, he adds, "This study is a data point. It is valuable for people concerned with the TCO. But it is not the only valuable data point."
Kusnetzky admitted that he was actually surprised that in the study, Linux was only marginally more expensive than Windows 2000. Based on other surveys of this type, he was expecting a much greater difference.
One of the other meta-trends that IDC has noted over the years is that the more centralized the product, the less expensive it will be in TCO. Given the de-centralized nature of Linux, Kusnetzky expected a much higher cost result for the open-source operating system.
As it turned out, in network service workloads, Linux was just 11% more expensive, and in the area of security, 22% more expensive. In Web services, Linux was actually 6% less expensive.
Based on earlier trends, "I was expecting Linux to be orders of magnitude more expensive," Kusnetzky said. That it wasn't, he added, "indicates that the Linux and Open Source community has done an excellent job putting tools together."
The survey's timing is also significant, Kusnetzky said. The survey was done last summer, before the advent of UnitedLinux and Red Hat Advanced Server. Done today, one wonders how the survey results would have differed.
As mentioned earlier, Kusnetzky is aware of the sharp criticism this report has received in the Linux community. But, after the initial round of complaints, he is counting on one of the communitys great strengths to kick in. "Most people [in the community]," he said, "will take the information and make their products better."
And, in the Linux arena, things can be made better very quickly, given the sheer size of Linux's development team versus a company like Microsoft.