LWN.net Logo

Linux and the total cost of ownership

We have seen, in recent days, a flurry of reports and analyst proclamations to the effect that, while costing more up front, Windows ends up being cheaper than Linux when the "total cost of ownership" (TCO) is figured. This cost includes things like staff time, training costs, etc. Certainly it makes sense to take a broad view of what a particular computing system really costs to operate. And, certainly, the analyst reports are objective; they would never, ever, after all, bias their reports in favor of the large corporation that has paid for the work.

Even so, some questions come to mind.

Your editor, who, in a previous life, managed a medium-size system administration group, observed that a single Linux or Unix system administrator could handle about twice as many systems as a single Windows administrator. As Windows systems replaced Unix systems on desktops, the administration staff had to grow. Many others have publicly noted a similar pattern. The observations of people actually running system management groups do not carry the weight of a scientific analyst report printed on Very Heavy Paper, but one might still ask: how is it that Windows is cheaper to run when more people are required to do the job?

Windows systems have well-known virus problems. Large scale virus attacks have led to direct costs for companies estimated in billions of dollars. Most large networks require constantly-updated virus scanning systems, active mail filtering, and regular "don't open that attachment" user cluestick sessions. All this is expensive; have these costs been figured into the TCO calculations?

Amazon.com claims to have saved $17 million by switching to Linux. E*Trade, too, saved a lot of money by going to Linux. The City of Largo, Florida, claims to save at least $1 million each year from its switch to Linux desktop systems. Why didn't they switch to Windows, if it is so much cheaper? (As an aside, this NewsForge followup on Largo is well worth a read).

Linux-based systems can often run on the same hardware, without upgrades, for longer. There is far less pressure for constant system upgrades - and no EULAs requiring such upgrades. Have the costs of the additional hardware and software upgrades required by Windows been taken into account?

Software license management is expensive. Companies must track the license for every application installed on every system on their networks, and they must cope with occasional annoyances like BSA audits and raids. Tracking thousands of licenses on thousands of systems is not a part-time job; have licensing compliance costs been figured into the TCO studies?

And so on. The real point is this: we should not give up the TCO argument easily. Linux systems are, beyond doubt, overly difficult to administer - especially for certain kinds of environments. There is a lot that can be done to reduce ownership costs for Linux systems. But, even so, the "Windows is cheaper" argument has not been made in any sort of convincing way.


(Log in to post comments)

GNU/Linux often has a lower Total Cost of Operation

Posted Dec 12, 2002 13:36 UTC (Thu) by ber (subscriber, #2142) [Link]

David Wheeler's frequently updated article Why Open Source Software / Free Software (OSS/FS)? Look at the Numbers! looks at all studies and draws the following conclusion (among others):

Total cost of ownership for OSS/FS is often far less than proprietary software, especially as the number of platforms increases. These statements are not merely opinions; these effects can be shown quantitatively, using a wide variety of measures.

GNU/Linux often has a lower Total Cost of Operation

Posted Dec 12, 2002 14:18 UTC (Thu) by jquinn (guest, #605) [Link]

While I've seen a lot of people talk about how Windows systems require an MCSE to run them, I wonder if they've factored in the rather significant cost of that. While some people may pay the $5-10K themselves, it's been my experience that most companies pay it for their people.

They said that OSS requires an expensive person to run. Expensive, because such people are rare, although that's changing quickly. They also say that OSS administrators have to be smarter. If they keep on saying OSS admins are smarter and better paid, I see a mass exodus of admins from Windows to OSS.

Windows and the total cost of ownership

Posted Dec 12, 2002 15:38 UTC (Thu) by dbreakey (guest, #1381) [Link]

Windows is "cheaper" to run for several reasons:

  • Inertia. It's worked well enough for us up to now; why change?
  • Top-down pressure. The company just got this sweet deal! Two thousand licenses at over 25% off!! The sales guy said we couldn't get a better deal anywhere else!!
  • Standards compliance. Microsoft is such a big company, sure, but even a huge company wouldn't go against Industry Standards, would it?

The problem seems to be that there are just too many executives who refuse to look at real world figures, which tend to be messy and disorganized; those nice, neat figures in the latest analyst's report are so much easier to deal with. And those analyst's customers are us, right? After all, we're the ones who paid a premium to get the report…

Fortunately, those seem to be a dying breed. Slowly, to be sure, but still dying off.

Sept. 2002 MS strategic review

Posted Dec 12, 2002 15:54 UTC (Thu) by jamienk (guest, #1144) [Link]

One aspect of this that's been getting little play is that in the latest leaked MS Linux stratagy memo, MS telegraphed this move.

...we should not expect rational arguments focused on undermining support for OSS, Linux and the GPL to perform well. In the short term, then, Microsoft should avoid criticizing OSS and Linux directly, continue to develop and aim to eventually win the TCO argument...

Windows could be cheaper

Posted Dec 12, 2002 16:16 UTC (Thu) by dbrandon (guest, #1090) [Link]

The irony of this is that OSS desktops probably are cheaper, but it doesn't have to be that way. For example, we have a lot of Windows 2K desktops here. How do we keep TCO down? For example we don't allow people to be admins of their own machines. Most companies, in my experience, don't do this, which I find baffling.

Also, Windows has a pretty good, fine-grained security model -- which is almost completely unused. Almost every MS and 3rd party service runs as SYSTEM. (Everything running as root is an issue on Linux as well, but there are rumblings about fixing this).

So, on most Windows desktops, you have everyone logged in in the Administrator group, and every service running as SYSTEM. Of COURSE you're going to have problems. Every bug and virus is going to be a catastrophe. You could argue that default Unix security isn't as capable, but at least it's used.

Windows could be cheaper

Posted Dec 12, 2002 18:21 UTC (Thu) by esh (subscriber, #140) [Link]

Running a system process as "root" is often less of a problem than it is made out to be; and this is different from having set-UID root programs, which are more delicate. Keeping a system simple is often decisive for maintaining a secure system. ACLs, the basis for the fine grained security settings of Windows NT/2000/XP, has been available for most kinds of Unix system for many years. It is however only rarely used because it makes a system more complex. In most real world cases increased complexity means weaker security.

Incidentially, reducing complexity also reduces the TCO. The significant cost of MSCE training is justified by the necessity to learn all those concepts. OSS/free software on the other hand has many independent components each of which is usually easier to fully understand compared with the corresponding "component" in Windows. This is a major reason why OSS/free software is cheaper to maintain.

Windows could be cheaper

Posted Dec 12, 2002 22:09 UTC (Thu) by dbrandon (guest, #1090) [Link]

It's true that simplicity is key to keeping down TCO. But there is a tradeoff here. People harp on security a lot, but rarely really do much. Linux has not only ACLs, but also capabilities, which really help security wise. But these make things even MORE complex, of course. And I agree that not knowing what you're doing leads to weaker security. But sometimes, for example my environment which handles data under FDA regulations, you absolutely require that complexity.

However, I don't think that's a Windows/Linux thing. They both can be configured for good security. My (not very well made) point was that some of the complaints about Windows come from a kind of entrenched bad practice from the DOS days, where everyone was a system administrator and every process was a root process. My background is a unix one, so I find that absurd, but it's standard practice in windows. Some end-user software we use here just plain won't run unless the user is an Administrator. Crazy.

I do agree that the simplicity of Linux is a big plus. Well, I wouldn't say simplicity so much as transparancy. Sometimes, on a windows systems, it's practically impossible to figure out what the system is doing and why. But if you deploy your windows desktops in a unixy way, you can minimize the effects of this. No admins, standard configurations, server side data storage, minimal well-known configuration -- this allows you to simply take away a badly behaved box and plop a new one in front of a user with minimal fuss. This is standard practice in the unix world. Every user in the company likes to think that they own their windows desktop. They don't. I do.

So for desktops, I think it's a toss up. Most user problems are things like, "where's the on switch?" For servers, I prefer the transparency of Linux, which I think makes simple, minimal configurations more possible. Windows has some better enterprise-wide tools, like MMC, which linux just plain can't touch. Overall, though, my experience is that linux tends to be somewhat more solid.

The whole windows/linux thing is kind of a red herring, though. Let's face it: software sucks, no matter the platform. The interfaces haven't advanced in 30 years. What's the substantive difference between the windows/X/Mac OS X interfaces? Why, nothing. (Might as well use the free one in that case, BTW.) My anecdotal experience suggests to me that TCO differences will be rounding errors over the lifetime of some piece of software. Right now, the biggest factors controlling TCO are process oriented. Training. Security. Coherent IT strategy. That's where the big wins are to be had. People are only now figuring out what personal computers are actually GOOD for. The whole process is in its infancy.

Copyright © 2002, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds