With Microsoft and Linux both claiming that they offer a lower total cost of ownership, IT users can be forgiven for being confused. Nicholas Enticknap cuts through the hype.
Total cost of ownership (TCO) is becoming the battleground on which the Windows-Linux war is being fought. Microsoft is running a marketing campaign designed to overcome the widespread perception that Linux is free and to prove that Windows costs less to run over a five-year period.
Microsoft has sponsored market research from prominent firms including IDC and Meta Group and published it on a new part of its website entitled "Get the facts". The headline of the section, "Leading companies and third-party analysts confirm it: Windows has a lower cost of ownership and outperforms Linux", underlines the point.
Microsoft's biggest weapon is the IDC report, "Windows 2000 versus Linux in enterprise computing," which argues that "Windows 2000 offers lower total cost than Linux in four of the five workloads common to most corporate IT environments." These workloads are network infrastructure, print serving, file serving, security applications and web serving. IDC concludes, "The cost advantages are significant: 11% to 22% less over a five-year period."
The reaction to Microsoft's initiative from the Linux community has been mixed. Among the die-hard enthusiasts there has been outrage, ranging from "of course Linux is cheaper" to questioning the independence of the market research. The IDC study was paid for by Microsoft.
From the industry side, there has been marked restraint. IBM, for example, which has a foot in both Microsoft and Linux camps, is being studiedly neutral. Regional Linux sales manager Andy Hoyles says, "If it makes sense to remain on Windows, or to go to Windows, we make that recommendation." But he adds, "It does not matter what anybody says, Linux is something our customers and other people's customers are getting because of the cost savings. That is the reality of the situation."
The Open Source Development Lab (OSDL), a body set up to promote the cause of enterprise-level Unix, says, "OSDL's position is that, at this moment in time, it is not a question of either Linux or Windows. There is a place for both, depending on requirements."
So are the Microsoft "facts" any help to users? Superficially, they are - TCO is a real issue. IT departments have to adhere to their budgets and lower costs, especially lower purchase prices, are a potent sales argument in a time of financial restraint.
It is clearly prudent for Microsoft to tackle the widespread perception that Linux is cheaper head on. But is IDC's claim that Windows servers are generally cheaper to run than Linux servers true?
Here we start to run into deep waters. TCO calculations are fraught with difficulty because they depend on so many assumptions. Each organisation needs to do its own sums and the results will vary depending upon the user's circumstances.
There are some basic facts which will hold true for all users. Linux hardware and software acquisition is cheaper. Hardware and software costs are a small part of the TCO - typically about 10% over five years. People costs are the largest element, accounting for between 50% and 66%, depending on whose study you read.
On the technical front, Microsoft products are more tightly integrated than Windows products, which makes the cost of management lower. Microsoft's head of platform strategy Nick McGrath says, "The TCO savings come from what we call 'integrated innovation'. All our software is designed to work together in an integrated fashion. As requirements increase, innovation allows you to easily extend the product."
Bloor Research senior analyst Tony Lock supports this view. "One thing is clear, in the management of any system, people and management is the bulk of the cost. That works in Microsoft's favour."
Costs also depend on the application and IDC has admitted this. For one of its five application categories - web serving - it concluded that Linux was cheaper. This is supported by a survey that can be found on the IBM website, produced by the Robert Frances Group (RFG), entitled "Total cost of ownership for Linux in the enterprise".
The study concluded that "Linux was the least expensive platform to deploy and operate" when compared with both Solaris and Windows on web page serving applications.
Linux licensing models are more flexible than those for Windows (or proprietary Unix), giving greater freedom of movement. This can provide bigger savings than are immediately apparent. The RFG study made the point that Microsoft's aggressive licensing policy has driven people to implement software inventory products at a cost ranging from £2.65 to £26.50 per seat. IDC has not included this cost in their calculations.
Linux runs on a number of architectures including Sun Sparc and IBM Power as well as Intel, giving the user more choice. What value do users place on that? For many, it will have no value - they will happily run whatever they buy on the same server architecture throughout the application's lifetime. But for some this freedom of movement may prove critical. At the very least, it is a bargaining counter which can help keep costs down.
Users also have to second guess the future. One reason why Windows systems take less management effort is because Windows is more mature than Linux. Windows is continuing to evolve - there is general agreement that the management capabilities of Windows 2000 are a vast improvement over those of NT - but Linux is evolving more quickly. More and more suppliers are porting products to Linux or developing products for it. As a result, the present day management costs for Linux will decline over time. By how much is a matter of guesswork.
Windows offers a wider choice of more established applications. This is really an on-off switch rather than a quantifiable cost component; it is only critical if you want one of those applications now. The same thinking underlies IBM's attempts to migrate Microsoft Office to Linux.
The RFG study says, "Historically, Microsoft has a poor reputation for product security, and although the supplier has been working hard to repair this image, RFG believes it will take quite some time to do so."
Microsoft has suffered a lot of adverse national media coverage recently because of damaging virus attacks, and also from the unauthorised publication of some of the Windows source code on the internet. Linux has a reputation of being much more secure, which derives from its Unix inheritance. The IDC report concludes, "For most workloads, Windows servers experienced higher downtime."
Security does translate into costs, but again it is very much a value judgement as to what they are. The RFG found that, "Survey data showed that Windows installations require twice the number of administrator hours on the average amount of time spent patching systems and dealing with other security-related issues."
A fundamental weakness of all TCO studies is that a large element of the costs are people costs, and these can rarely be calculated. Lock says, "The big problem with TCO is that the bulk is to do with how you cost people's time. There are few organisations with enough information to know how long anybody spends doing a particular function."
There are also unquantifiable strategic considerations, such as dependence on a single supplier. The controversial Microsoft licensing arrangements introduced last year have brought the impact of this home to many users (and the costs this involves are not quantified in the IDC report, which was prepared in 2002). David Roberts, chairman of the corporate IT forum Tif, has said corporates are looking at open source because of the perceived cost of Microsoft products and its dominant market position. Getting out from under such dominance was another reason why users deserted IBM in the early 1990s.
The arguments for open source are not as simple as proponents would have us believe. SCO's attempt to assert intellectual property rights over Linux highlights just one area of complexity. TCO is not the most important issue, any more than the purchase price is. In the last analysis, return on investment is more important.
But ROI is even more difficult to forecast than TCO is to calculate, and it is impossible to produce a report arguing that Windows produces a better ROI than Linux (or vice versa). The choice between Windows and Linux in the last analysis depends on where the user organisation is, where it is going and what it is trying to achieve.
Lock says, "Every business has to sit down and look at its IT and say, 'What do we want to do as a business? How do we get from today to tomorrow? For some, the right answer will be to stay where they are, and others may move forward. You have to do the groundwork."
The TCO argument is not really about the facts - it is a marketing battle. A perception had arisen that Linux is free and Microsoft is trying to show that it is not.
Whatever the merits of the focus on TCO, it can be seen as a sign of the maturity of the IT industry. In the past, arguments between competing suppliers and platforms have usually majored on performance, and the facts have been in the form of benchmark data. At least now the battle is being fought over an issue much closer to the boardroom's heart.
IBM's marketing war
For IBM the Linux/Windows battle has echoes of the early 1990s, when it was fighting a similar war with Unix. Mainframes were then perceived as being more expensive than Unix servers and their prices were undoubtedly much higher. IBM did not argue the point, other than stating that running costs per user and per transaction were markedly lower. But it was a time of recession and users were not impressed. They "downsized", in the jargon of the time - that is, they went out and bought Unix servers in droves.
The outcome was calamitous for IBM. It posted what was at the time a world record annual financial loss for any company. Heads rolled in the boardroom and an outsider was brought in to take charge for the first time ever. IBM lost its hegemony of the computer market and it has never got it back.
This article is part of Computer Weekly's Special Report on Linux produced in association with Novell
This was first published in March 2004