Part II: Corp. Desktop Linux - The Hard Truth

comforteagle writes "I've published Part II of W. McDonald Buck's essay on Linux TCO. In it he looks at the scenario of a company having already moved to Linux in the server room and also to open source software on Windows desktops, but "...now wants to know, how much extra can be saved by the final step of changing the operating system itself? And, what are the other costs, risks and benefits of doing that. To keep the scenario simple, we're assuming too that this will be done at a time when the desktop equipment is also being replaced. The news is good, but not as good as we like to believe." Part I was discussed previously on Slashdot."

Comments

Article text, in case of slashdotting

[Anonymous Coward]

Article text, in case of slashdotting:

Part II: Corporate Desktop Linux - The Hard Truth

Enterprise Linux / Article
Date: Feb 11, 2005 - 06:39 AM
by W. McDonald Buck (slashdotting-resistant version)

In Part I (of IV) of Corporate Desktop Linux: The CTO's Hard Truth retired CTO of World Bank W. McDonald Buck wrote about the difficulty of making accurate real world assessments of Linux vs Windows TCO analyses: If Windows laden computers are less expensive or equal vs. no OS in cost for Joe Hacker.. where's the savings for a large buyer who gets a volume discount? In this, Part II, he addresses the hard cost savings for companies of switching the desktop operating system, as distinct from the applications, as he sees it.

Part II: The Hard Cost Savings of Corporate Desktop Linux
A number of cost studies have been published. In addition to their biases and their other flaws, they generally evaluate replacing both servers and desktops with Linux. There is no question that Linux is successful in the server market, where almost any fair analysis concludes it will save money. This article focuses exclusively on the desktop side of the equation, since for a decision maker, these can easily be separated into separate business decisions.

A different way of framing the question will help get at this. Say a smart company makes the decision to save a lot of money by moving all its servers to Linux. Many have begun doing so. And of course you don't have to change operating systems to save money on applications. The press is full of reports of people dumping Internet Explorer for Firefox. Switching from Office to OpenOffice for all but the most demanding users is a smart move with a high payoff. Having done those things, the company now wants to know, how much extra can be saved by the final step of changing the operating system itself? And, what are the other costs, risks and benefits of doing that. To keep the scenario simple, we're assuming too that this will be done at a time when the desktop equipment is also being replaced.

The news is good, but not as good as we like to believe.

We know that Windows costs money, and we believe that overall, a Linux desktop must cost less. If this is so, why is the cost benefit analysis unable to sway more migration on the desktop? Part of the reason is that switching the operating system is a more drastic step, precisely because it forces the migration of all the applications at the same time, increasing the switching costs. The hard truth is that for the majority of the desktop market, the capturable, present value cost advantage to the operating system change (as distinguished from the other changes) doesn't seem to outweigh the costs and risks of doing it, unless you put a high value on the benefits of Linux. Those benefits, which are so compelling to us, are not compelling to the target audience. This article explores the hard dollar cost side of this heretical assertion. A follow-up article looks at the benefit side.

Before getting to the analysis, I want to also briefly pay homage to the idea of changing the whole computing model, as many suggest, to more of a thin-client central server architecture. I think companies will increasingly do this, but corporate environments change slowly, and for many of them this is a big change. In large part this whole series is about pressing for the winnable victories first, rather than the more radical changes. Since thin-client looks like a more radical change to the corporate world, I'm going to stick with analyzing the less radical move, acknowledging that the other may be a big win if it can be sold.

As a point of reference I'll use the 2004 update of the TCO analysis done by Cybersource Pty. Ltd, entitled Linux vs. Windows, Total Cost of Ownership Comparison (.pdf). This is a very pro-Linux report (egregiously so in my view). Unfortunately, I found a number of errors (11), including 5 pretty obvious arithmetic or transcription errors. Still it has the virtues of trying t

Honest question

[Gob Blesh It]

I didn't start looking into this recently. But is it really impossible to: (1) encode MP3s and AACs, (2) render text using TrueType kerning tables, or (3) take advantage of subpixel rendering, on Linux desktops, without breaking patent laws? With all the hype surrounding Linux on the desktop, I have trouble believing these common activities could be illegal, but nor have I seen anything indicating otherwise. What's the deal?

(And please, no Stallmanesque rants about how intellectual property should be abolished. I don't have the patience for these ill-conceived prescriptive arguments.)

Re:Honest question

[bersl2]

I know that #2 is certainly false. Freetype2 uses autohinting by default, which is not covered by the Apple(?) TrueType patent.

Re:Honest question

[ratboy666]

Yes, but violation of patents has NOTHING to do with what platform you are on...

Eg. MP3 usage. Patent is "free", even for commercial programs, until you sell 10,000 units a year. At that point, it will cost you. Doesn't matter whether is Windows or Linux, or XXX.

Ratboy

Re:Honest question

[msaavedra]

And please, no Stallmanesque rants about how intellectual property should be abolished.

Are you implying that Richard Stallman has said that intellectual property should be abolished? I doubt he would say that, considering that he doesn't even think "intellectual property" is a useful term, since copyright, patents, trademarks, trade secrets, etc. are all separate concepts governed by separate laws and having varying effects on an individual's freedom.

Perhaps this is not what you meant, but I wouldn't be surprised. It seems that every day here on /. someone inaccurately attributes all sorts of loony beliefs to Stallman. Most of his opinions are fairly cogent and reasonable. Of course, he doesn't help himself, with his unorthodox appearance and behavior, lack of social skills, and pedantry regarding unimportant topics like the whole GNU/Linux thing. Still, unlike Joe Sixpack, I would think that the typical geek would be more tolerant of such things.

Anyway, to keep this post on topic:

1. MP3 is definitely patent-encumbered, and I believe AAC is as well. For this reason, Fedora and probably other distros do not include MP3 players/encoders. I don't know if using a free encoder in and of itself is breaking patent law, since the patent holders have not indicated any wish to forbid this (at least in Fraunhofer's case). After all, patent holders aren't required to charge you a licensing fee. Also, there is nothing specific about Linux that forbids using licensed apps; I believe there are some proprietary media apps for Linux that have purchased licenses for MP3.
2. There have been some issues with Freetype's bytecode interpreter in the past violating an Apple patent. I'm not sure if this has anything to do with kerning or not; I seem to remember it dealing with hinting. The last I heard, the Freetype developers had worked out some new ideas that don't violate the patent and gave better results anyway. On my Fedora desktop (and Fedora takes patents very seriously), the fonts look excellent, much better than on Windows. OS X may still have it beat, though.
3. I've heard nothing about patents on subpixel rendering one way or the other. I know that it works on my Fedora desktop, and I can't imagine they would enable it if it violated a patent, considering their stance on other patent-encumbered technology.

Kind of an old issue.

[fm6]

The TCO issue is kind of worn out by now. For years now, Linux advocates (and before them, the advocates of network computing) have tried to convince IT decision makers that they needed to get their TCO down by moving away from Windows. I've always thought the argument was a pretty good one, but it's never been convincing to the decision makers, who just haven't been willing to make the necessary paradigm shift.

Microsoft's current inability to handle security issues is much more persuasive. Linux advocates should focus on that, instead of beating a dead horse.

Re:Kind of an old issue.

[Noksagt]

Microsoft's current inability to handle security issues is much more persuasive. Linux advocates should focus on that.

Only if they want to continue to be several steps behind. Most still consider the security to be "good enough." Indeed, the "good enough" argument makes it very difficult to win most arguments on the technical merits of Linux over MS. MS is also making very public moves to improve the security of Windows. You really do have to attack their model & try to force a paradigm shift, rather than to highlight some issues that will eventually be resolved.

Re:Kind of an old issue.

[kbielefe]

I think you hit the nail on the head about TCO. Any switch is going to be painful and expensive in the short term, making TCO comparison only useful in the long term. The problem is TCO is extremely difficult to predict in the long term, especially for a highly customizable environment like Linux.

I also think that most people that use Linux didn't originally switch because of cost. For myself, I put Linux on a spare laptop when I was doing contract cgi programming and wanted to be able to work offline. It was many years of off-and-on experimenting with various distros and purposely buying hardware that would be Linux compatible before I permanently erased my Windows partitions, and a couple of years after that before it really saved me any money. I suspect most Linux advocates have similar stories.

I don't really think focusing on security is going to do it either. Microsoft is making daily improvements in that department. I believe that focusing on reducing the switching cost will have the most benefit. When the short term cost is low, it is easier to see the long term cost reduction.

What does that mean? Advocate choosing OS and vendor neutral solutions when your company is making a change anyway. Advocate creating documents in open formats and do it yourself. Help make free software easier to install and configure. Help make Linux easier for the Windows administrator or CTO to install at home and "take for a spin."

Re:Kind of an old issue.

[fm6]

I don't really think focusing on security is going to do it either. Microsoft is making daily improvements in that department.

No, they're making more noise about security. As far as I can see, they're actually losing ground, even though they're throwing more and more resources at the probem.

Go thin.

[QuietRiot]

One way to aproach this situation is to put a large multi-user box in each individual department or workgroup. Keep the windows desktops there, add X-servers, and run some apps centrally.

If the question is licensing, a net-booted corporate (Linux/FreeBSD/NetBSD) desktop wouldn't be too difficult. Run locally what the machine can handle, make sure the network is super fast, and run the apps requiring performance on the workgroup server.

This would reduce the need to upgrade lots of hardware (which will surely be obsolete again soon) and minimize downtime in the process. If need be, some legacy apps could survive under Wine locally or, again, at the workgroup server. Keep storage centralized to facilitate backups across the entire organization (OpenAFS?).

It's an old way of doing things but overlooked far too often. You've obviously got to run the numbers but, surely, "a few good men" handling things on the server (they would handle the app server AND the centralized, consistent-across-the-organization, netboot image(s)) would be much less expensive than the workload on an IT staff required by an office full of people and their problems on Windows machines.

Maybe not *THE* solution, but certainly worth a look for many.

Pixelate

[QuietRiot]

It did seem offtopic at first... but...

Out of the 3 you mentioned, subpixel rendering is actually the one I was thinking could prove most useful in a corporate environment. It *does* help, and with the number of LCDs making their way onto desktops today, this feature would benefit free systems (free system users) in a big way.

Whether or not the boss will take your eyestrain into consideration when choosing a business platform is another question altogether.... :)

Re:Pixelate

[paulatz]

Gnome have a "subpixel antialias" or somthing like that in the fonts preferences.

Leaving Windows is probably not a huge TCO savings

[LWATCDR]

Right now you are not likely to save a huge amount of money dropping Windows. The big savings will come from dropping Office for OpenOffice, dropping Outlook for Thunderbird and IE for Firefox. Once you are not tied to Windows only apps then you are free to move to an other OS easily. Be it Linux or Mac.
It is a matter of freedom.

The real question one should be asking is

[xenocide2]

How much and in what ways can a switch improve employee productivity? The author makes a valid point that the 100k saved is minimal in transitioning the OS, but anyone seriously looking to promote Linux from within must be prepared with a host of reasons that translate into a more productive employee. Saving money on IT operations is one thing, but one can be quick to lose sight of the function of computers in the workplace-- make smarter decisions faster.

I get the suspicion that part IV will be the article that discusses this.

Re:The real question one should be asking is

[Vengeful weenie]

I'll give you one very good benefit. The ability to keep users from installing malware, trojans and spyware. Keeping the users out of the system areas is more than enough reason, and will save tons of maintenance time.

We're keeping a close eye

[mnmn]

We're about 80 machines including 10-ish servers, almost all MS except the firewall thats OpenBSD, and a specialized Unixware box that cant be touched. Two apps are really keeping us from a 100% Linux rollout, Lotus Notes and the ERP software. The ERP company has promised linux binaries, and we've seen some test cases, but thats about it. Please dont ask us about iNotes, we use Notes far too heavily to attempt that (and we have).

So we're just keeping a hopeful eye out, especially on Lotus Notes, and the desktop distros. The server bases are covered, between SuSE, samba and sybase, a majority of our operations can be moved. What we want from the desktop distros is that it should look and feel a LOT like the windows NT/2000/XP interface. Notes shouldnt be hard to port at all, its all java based anyway, IBM is dragging its feet for linux there, despite some customers asking for it.

We'll probably not be the first to go all Linux (or FreeBSD, or BeOS, or SkyOS, or ReactOS etc), but its interesting to know how far in the future we can expect it.

Complaints:
(1) Theres no single Linux distro that really looks and feels like win32. Training dozens of people to use a different interface from what theyve been using for 10 years, is something we cant stomach. Tried Fedora3 and Xandros, will try Linspire soon.

(2) Application seriousness. Everything is made for win32 first, betas appear for linux and stay beta. That doesnt fly with corporate networks. We need a FEW critical app vendors to really support Linux, once the market floats, the rest will follow suit.

(3) $$$ and energy. It takes a LOT to switch ERP systems. Takes a LOT to replace network system. Takes far more to replace everything. Sure we can do it in steps, but collectively the steps are all manhours to attain OSS nirvana, but not profit in the short term. We cant just aim for everything free, somehow, in some ways, theres still a difference between debian and readhat. things get patched in redhat and developers have to listen and work. With debian, the design and philosophy is awesome, but you cant really ASK a developer to fix that bug that is stopping your company. He'll just say fix it yourself.

(4) Assurance. Not an issue since our mgmnt hates MS, have hated it since the DEC days (which is why we ended up with lotus instead of exchange).

So we're all getting closer to the 'threshold', still not there yet.

One important point is missed here.

[shaitand]

The author talks about the OS transition being "risky". In what way?

The servers and desktop OS are changes that the user is not much concerned with. It is changing the applications which impacts the users. Once you have changed the application set to one that is not windows dependent I have trouble believing that changing the underlying OS would have much impact at all. Basically being able to get rid of the OS fees is an inherient benefit to getting rid of OS dependent applications. After that the workstation OS can be changed at will without significant retraining of users.

Like the servers, the users do not interact with the OS directly, only with the application icons on their desktop.