IDC: "People Costs" Make Linux Costlier Than Windows
When it comes to comparing costs between Linux and Windows 2000, a study from International Data Corp. has found the real expense is not the cost of the software -- but the outlay needed to keep those servers up and running.
In fact, IDC determined that the cost of hardware and software for either Windows 2000 or Linux will run less than 10% of the total cost of ownership over five years. The tools and the trained staff a company needs to keep servers online -- and the expense to productivity when servers fail -- make up more than 80% of costs -- the vast majority of server expense, IDC found. The study also noted expenses arising from "life-cycle costs," including strategic IT choices, company standards, applications availability, deployment and performance considerations.
In specific, the study compared Linux to Windows 2000 servers in what IDC called "five common workload scenarios" -- network infrastructure, print servers, file server, security and web servers.
Over five years, IDC found that the costs for Windows were 11-22% less than for similar Linux servers in four of the five categories (network infrastructure, print servers, file server, security). In web serving, IDC found Linux had a cost savings over Windows 2000 of 6% over the five-year period. IDC calculated TCO based on six key areas of expense: Hardware (4.4%); Software (4.6%); Outsourcing (0.4%), Training (5.3%); Downtime (23.1%) and IT Staffing (62.2%).
But, IDC found IT staffing "by far the most significant cost [area]." In this category of tools and trained staff, Windows 2000 maintains an advantage over Linux, IDC noted. In fact, in today's market, IDC found that in some cases, the availability of Windows 2000 trained staff can cut personnel expenses by 50% over some Linux staffing requirements. The exact numbers, based on the study's methodology (see below), put staffing costs at $60,000 for Linux servers versus $40,000 for Windows 2000 servers over five years.
The small impact of hardware and software costs over five years -- and the big-ticket expense for IT staff -- was the most noteworthy finding, Al Gillen, IDC's longtime research director for systems software, and head of the study team, told IDN. "A Windows 2000 license for 500 users over five years can cost as little as $2,500 per year -- hardware and software included, which is not much of a difference over 'free' Linux software with hardware. The real costs are from keeping each of those servers up and running."
Windows and Linux -- a Microsoft Perspective
To get a better perspective on whether Microsoft intends to have Windows NT work with Linux, not simply against it, IDN spoke with Peter Houston, senior director in Microsoft's Windows server product management group.
IDN: IDC mentions the area of network/systems management tools, and that Microsoft is not "standing still" Are there key application-level areas of management where Microsoft feels there are key advantages in Windows NT/2000 (especially for developers building /NET applications)?
Houston: Through the addition of key management services at the heart of Windows and .NET, and a range of highly scaleable value-add management products, Microsoft is dedicated to making Windows the most manageable platform within a heterogeneous enterprise.
Recently, Microsoft has taken Windows management one step further, through the availability of the Microsoft Solutions for Management. These solutions provide customers with blueprints for addressing key management issues for achieving operational excellence through the use of standards-based best-practices used in conjunction with Windows and the management technologies above. [Developers can get an update on management at Microsoft Management Main Site
IDN: During last year's LinuxWorld, Microsoft attended for the first time, and announced several add-ons to NT that would make it easier for NT to work with, and manage in an end-to-end way, NT-to-Linux connections. At the time, Microsoft also said this was a "first step" to embracing Linux servers in the enterprise with Windows NT and/or Windows 2000. With LinuxWorld coming again next month, what other Windows-to-Linux integration features (for management, security, etc.) will Microsoft offer?
Houston: First, I just want to clarify that the product is Windows  server, not NT. And, while I can't comment on any specific future plans, we are always listening to customers and taking feedback.
One [recent] example of this is Microsoft's Services for UNIX3.0 (SFU), which provides supported and fully integrated cross-platform network services geared towards enterprise customers who want to integrate Windows into their existing UNIX environment. SFU 3.0 helps meet our customers' needs of reducing costs by allowing them to work across platforms.
IDN: IDC's study also mentions that Windows 2000 does better than Linux in workload issues. This was a bit confusing to me, given IBM has ported DB2 and Oracle has ported its high-end database apps to Linux. Can you clarify this?
Houston: The IDC report uses the term 'workload' to refer to the specific roles in which a server can be deployed. So, for example, a web server is a workload, as would be a file server. The IDC study examined 5 common (and relatively simple) workloads (web, file, print, network infrastructure and security services). They didn't study 'database server' workloads or 'application server' workloads.
That said, we believe that the greater the complexity of the workload, the greater the TCO advantage Windows will deliver over Linux. For example, if you were to compare Windows to Linux on databases, you would need to purchase something like DB2 or Oracle (they are not free on Linux), SQL Server will be less expensive for Windows, and the cost of the OS itself will be a small fraction of what you need to buy.
Inside the Study's Methodology
The IDC study, which IDC execs said was sponsored by Microsoft Corp., has generated some skeptical and dismissive reaction from the Linux/Open Source community.
In light of that potential conflict-of-interest, Integration Developer News spoke with Gillen to get his reaction to charges from the Open Source community that the IDC study results were influenced by Microsoft.
"We were fully aware of all the negative response, and that had to be expected. But IDC has rules of engagement, and we used the same methodology we used with a study we did last year [on Linux usage] for Red Hat," Gillen told IDN. "We have rules of engagement for every study we do, vendor-sponsored or not. And if we do a vendor-commissioned study, it shows up right on the front page of the study so everyone knows it."
"We didn't intend to show where somebody won and somebody lost," Gillan stated. "Those stories miss the point. Our methodology is the same no matter who our client is."
In specific, Gillan said that Microsoft was not allowed to pose questions, propose companies to survey or handle the results. "Microsoft was only allowed to suggest which platform from their side would be compared," Gillan told IDN. "All the areas of comparison and the questions came from our existing system software research, which is why we had so many [IDC] members on the team," he added. "The point was to provide data and charts that will show where the costs are during an entire life cycle. We simply started out to help customers to get a sense of where the costs, not [to claim that] somebody won or lost by x%."
Gillan also told IDN: "We tried to avoid any religious biases on either side [Windows or Linux], Gillen said. "In fact, 40% of respondents told IDC that they used both Windows 2000 and Linux in their companies." To avoid any Windows bias that might have arisen from asking questions about "mission-critical" installations, IDC's staff did not ask about transactional environments, such as financial transactions processing, Gillan added.