by W. McDonald Buck, Retired CTO of World Bank

The past two articles [Part I | Part II of IV] discussed the hard dollar cost savings from switching desktops to Linux, particularly business desktops. Linux has no license fee while a Windows license nominally costs a bundle. This has to be a cost advantage if the cost of Windows licenses can be saved. Obviously this isn't the case with existing equipment, but even for new equipment, saving a lot of hard money by switching the operating system is not as big an instant winner as switching the servers and the applications. Still, we calculated that a company with 250 seats could save almost $115,000 over 3 years that was directly attributable to a switch to Linux. If this is free money, a company would be foolish not to grab it, even if there weren't additional benefits to Linux (such as better security and stability). But before switching from the cost side, we need to explore whether, from the point of view of the corporate CIO, there are additional costs – hard or soft – from a switch to Linux.

Part III: The Hard Truth About Linux on the Desktop – Soft Costs

You are the CIO of a small to medium sized company. You want a big splashy cost-saving victory to impress your board. Consider your thought process as you evaluate whether a desktop switch to Linux can make you a hero without getting you toasted.

Remember that you have already resolved to achieve savings by moving the servers to Linux where you can, and moving to open-source applications. You can start by migrating the browser, declaring truthfully that you've protected the organization from various security risks, though not saving any hard money. That done, you resolve to replace Office with OpenOffice on most of your Windows boxes at pretty low risk, actually saving real money.

But what about changing the operating system itself? Recall, the scenario is that you are changing hardware anyway, since changing operating systems on in-situ desktops is a high cost operation even if it is license free, even if it were Windows to Windows. To keep the analysis simple, we get to do this at hardware change.

You realize that if you embark on this course, and fail, and have to go back to Windows after all, the transition is going to cost more, a lot more. Another way of saying this is you have to be committed, you can't do this halfway. If you get off the "conventional" path, and have to get back on it, it is going to cost a lot, and will certainly get you toasted. This is a significant risk.
And you realize that when you switch to Linux, you have to have a plan for every single Windows application. While you can switch to open source applications one at a time over a protracted period, switching the operating system requires changing all the applications at the same time. Office, browser and terminal apps are all relatively easy (and low cost). Legacy Windows apps and little databases, and a host of other such stuff will have to be converted, or emulated (eg Wine), or centrally hosted (eg Citrix). Take the scheduling software your boss and his secretary use, or the CFO's palm pilot. None of these are impossible, but every one of them adds to your costs and eats into your $115,000. If you have 10 or 20 or even 50 of these to convert, and a few of them prove tough, you could easily spend 10 staff weeks on it, or a lot more. What is that for your company in fully loaded (salaries, benefits and overheads) dollars, something in the range of $20,000? Moreover, each one of these increases your risks, and adds a potential surprise toasting.

Mail is easy if you happen to use standards based mail, but many don't. If your company has fallen into the trap of using a mail system with tight integration between the server and a native client app, like Notes, you get a lot of seductive user functionality for simple applications, forms and routing (and IBM/Lotus gets a really great lock in device). If you use one of these systems and switch the desktop OS, you must either migrate the mail system (very costly and difficult), or you must abandon the native client and with it some of that wonderful functionality, and probably some of the apps. Sure, there are some products that purport to ease this transition. But however you do it, migrating a tightly integrated mail client is going to be messy and costly and training intensive, and has many complex nits. If there is one thing that senior executives in the company rely on, it is mail; so you had better be sure you have a good answer for every nit, or you will surely be toasted. And again, every dollar you spend on making the transition comes out of the $115,000. Mail migration by itself could be a deal breaker, so let's assume you are lucky, and don't have to confront it.

Next, you will have to fund retraining costs. Trying to evaluate how much this costs yields a lot of conflicting information: some claim it is easy – very intuitive, perhaps easier than moving to a new version of Windows. Some claim that switching to a new operating system is a major league big deal. No doubt the truth is that it varies a lot by individual and averages somewhere in the middle. It doesn't really matter who is right here. The problem is that you, our hypothetical CIO, can't afford to under-estimate this. You must assume the higher costs and do the training. There will be problems in this switchover, as in any switchover; and if you did not do training there is a toaster in your future. How much does it cost to do even 1 hour of training for 250 people? The fully loaded staff time alone has to be $20,000 - $30,000 or more, and two hours of training doubles the figure. Then you add the costs of developing and administering the training, or buying it. I'll say conservatively $30,000 total, but use whatever figure you think is right for your company.

You may also have to retrain or rehire the IT support staff to prepare them to install and support the new environment, including all the things you had to add or change (Citrix, Wine, maybe changes in the mail infrastructure). Altering the support structure is very hard to put a defensible dollar cost on. Rather than pick a number out of the air, we hypothesize you have chosen your IT support staff wisely, and they are all already using Linux at home and enthusiastic about this good sense on your part. If you don't have this good fortune, you have to retrain (and backfill), or rehire (and overlap), increasing your costs, and your risks.

As a seasoned executive, you know there will be problems in any big cutover, so, you have to prepare the company for disruption during the switchover. An important high profile report will get delayed before a board meeting. A key customer proposal will miss the cutoff date. And of course, it doesn't matter if these glitches are due to the switchover; the switchover, and you, will be blamed. Although these costs won't be in your budget request, you know they will add to the overall company cost and each such event will cost you points. Too many points and you will get toasted.

Finally, you realize that since there have been very few major corporate cutovers from Windows to Linux, the press is starving for stories about companies attempting it. Your project, and therefore your company, is going to be under some scrutiny. This raises the stakes for your CEO and your board: the shareholders are watching. This in turn lowers the tolerance of the CEO and board for disruption and failure. Furthermore, if you fail so publicly you may not just lose your job, you may end your career.

What about cost benefits? One area we shouldn't ignore are the license management savings. Managing Microsoft licenses is a pain, and expensive. Dumping Office helps, but if you don't dump Windows you will still have to have someone look after the licenses. Still, avoiding that cost isn't a really big savings. If you are extremely generous, perhaps you could claim savings of 4 staff weeks a year. I wouldn't believe it, but somebody might. Whatever the real number, I feel safe saying this savings is offset by the costs we've assumed were 0 above.

This is the cost picture as you will have to explain it to the CFO. You can save $115,000 over three years, but you will surely have to spend at least $50,000 to $60,000 in the first year on the switchover to achieve it (and maybe a lot more if you revisit our generous assumptions). So, in the first year, this is not going to save money but cost money. You'll have to sell it as an investment. And, though the long term savings are appealing, this is an investment with significant risks.

Are the benefits, like stability and security, worth it? Perhaps. Next week we will have a look at the benefit side.

Corporate Desktop Linux - The Hard Truth is a four part essay: [Part I | Part II | Part III | Part IV]