The devastating case for Windows against Linux...

MICROSOFT HAS commissioned studies left and right to counter Linux's perceived cost advantages (See here, for example). These studies have been universally decried among Linux enthusiasts, and every flaw has been pointed out in one way or another. Rather than completely debunking the flaws, most of the critics simply point out that there is still a lot of work to do on the Linux front.

While no one ever said Linux was 'done', one of the main studies promoted by MS, a CxO level look at Linux TCO here is of interest. While the authors begrudgingly admit in the third paragraph that Linux has some advantages against Windows, they go on to pretty well debunk any Linux advantage if the big picture is taken into acount. Please overlook that the author, Laura DiDio is the same person who still touts SCO as having a devastating case against Linux. She makes an occasional sound point here. Bear with her, her work makes Yankee a lot of money.

That said, here are some of the better points brought up in the article. I personally feel that they are all quite valid. They all have no easy answer to put Linux on top, and I tried to think of some. Many of the other points brought up in the article have an asterisk attached to them, and those were left out on purpose. This is meant to be a summation of the strong wins of Windows over Linux that MS is promoting in its 'Get the facts' campaign. Zealotry and profits aside, there are some valid points here, which I'll summarise.

In the short term, Windows will simply cost less than Linux to run your organization on. There are a bunch of reasons for this, most of them involve people. Lets face it, as a percentage, very few people are trained on Linux, and only a few of them are worth their weight in twinkies, much less gold. This is not a condemnation of Linux admins, just the general bell curve of competence.

Because of their relative scarcity, good admins can charge more than an MCSE, so your cost to run a Linux server correctly will be higher than a comparable Windows server. Should you need support, the same thing applies, but hits home harder. A decent admin charges more, so a rent-an-admin will cost even more, as will a hired troubleshooter. Windows simply costs less to have someone muck with.

A similar problem rears its ugly head when it's time to upgrade. Going from Win95 to Win98 was easy, they were similar enough. You could even live with the 98 to 2000 migration, the programs were pretty similar. Sure, things like AD caused some heartache, but it was not an impossible move.

Linux is very different from Windows, and upgrading will be a major pain. Pain in this realm equates to dollars. Add in the scarcity of experienced Linux admins, and you have more pain, and more money flow. Not good once again.

Slightly related to people and differences is training. When you move to a new OS, even if it is similar on the surface, that can cause. Again, the 95 to 98 move wasn't bad. 98 to XP with something as simple as a new desktop brought major pain, confusion, and training costs to the secretary set. Linux is very different from 98. Chalk another win up for Win.

All of these problems including admin work, support, training, and ease of upgrades will lessen over time as familiarity with this 'brand new' OS creeps into the enterprise, but that is the future. For now, all the people problems, unless your organisation is staffed by IT workers who must be ahead of the curve, will cost you a lot.

The other half of the money pit is software. Not necessarily the cost of the OS itself, but all those other things you need to make it work in an enterprise. This is where the cost starts to add zeroes at an alarming rate.

Few software companies have ported their apps to Linux, but the number is growing. You can always run your old apps, some of them anyway, on Linux as is. Some may require emulator, VMs, or other hacks, but it probably can be done. The problem with this approach is that all the bells and whistles that come with Linux are conveniently ignored. Why did we upgrade again if we are just running the same old apps? Good question.

A partial answer for this is the purported greater security of Linux. While it has the potential to be more secure, you need a competent admin to lock it down. These are few and far between. While some of it may be do-able with a white paper, checklists, and best practices manuals, nothing beats hands on experience. Do you have anyone with years of hands on Linux experience on staff right now? Another point in the column for good old Windows, again, Linux can't compete on the years of experience front, and would you risk your enterprise on something so 'unproven'? I sure wouldn't.

The minor security gains for Linux are outweighed by the lack of management software. Most enterprises have a large, rather monolithic management suite to run their computers. If this suite has been ported to Linux, you are in luck. Most have not, and probably will not be anytime soon. Software like this is large, unwieldy and complicated, and failure in any way is not looked on lightly by big companies. Porting takes time and effort, so if your package is not on Linux, you are probably in for a long wait. Basically, Linux enterprise management software is a rare species indeed.

Does it get better for Linux? Nope. Custom code must be ported. If you are really lucky, you can just recompile your old apps, and they will work. Again this negates all the advantages of moving to Linux, but it is much cheaper. If you have to port, and you almost assuredly will, costs can be shatteringly high to start with, and go up from there. Sitting back and staying with Windows looks better and better now doesn't it?

Then there is integration. Your new Linux boxes may not play well with the hardware you have in place. OK, some may work like a charm, Samba and other things that emulate the older protocols will help, but again, you are running the older protocols. If it works, and again, that is a big if, you may encounter weird bugs and assorted headaches. That SAN probably wasn't Linux certified when you bought it, right? You can't just slap a sticker on it and make it compatible. If you are lucky, there will be a BIOS upgrade to do the job. Unlucky, and this is much more likely, and you need to rip and replace vast swaths of hardware. This is something no CIO looks at as anything less than a career limiting move. Windows, the safe corporate bet, Linux, the not so safe bet.

So, people, software and hardware are all clear wins for good old Windows over Linux. Except for the really rare cases, Linux simply costs more. Live with it people. There is a huge, nearly unending support mechanism for Windows, hardware, software and people, all at the beck and call of your chequebook. In the spotty cases where it exists now for Linux, it will take a few more pages from that chequebook before it's right.

As the Microsoft promoted study rightly points out, few people will do a real TCO study for the cost of migrating from Windows to Linux. Less than half of you believe the Yankee numbers. Most will get the proverbial bug up their ass, and jump in. By the time they realize how much Linux is costing them it is too late.

If you deploy Linux without a good reason, it will cost much more, be a huge strain on your IT staff, and potentially compromise security. General staff will incur slowdowns and productivity loss until training occurs, and even then, won't be nearly as proficient on Linux as the old Windows for a long time. If you need help, flip a coin, and if it comes up heads, you may find someone that can help you. If not, where do you turn if there is no one familiar with your OS?

The big picture shows that Linux will make you pay, and pay, and pay, and pay. Sticking with good old Microsoft Windows, be it 2000 or XP, you know what to do, who to go to, and what it will cost. Familiarity breeds content here. If you are going to move to Linux, do two things first. Run the numbers, and I mean really run the numbers, don't just compare the initial sticker prices. Then find someone who can support you to the level that you have already. If a minor miracle occurs, you will be ready to move, if not, don't.

OK, that was pretty damning for Linux, eh? Now, cut the above text out and put it in the text editor of your choice. Do a search and replace, substitute 'Longhorn' for 'Linux'. You get the following little potent argument.

In the short term, Windows will simply cost less than Longhorn to run your organization on. There are a bunch of reasons for this, most of them involve people. Lets face it, as a percentage, very few people are trained on Longhorn, and only a few of them are worth their weight in twinkies, much less gold. This is not a condemnation of Longhorn admins, just the general bell curve of competence.

Because of their relative scarcity, good admins can charge more than an MCSE, so your cost to run a Longhorn server correctly will be higher than a comparable Windows server. Should you need support, the same thing applies, but hits home harder. A decent admin charges more, so a rent-an-admin will cost even more, as will a hired troubleshooter. Windows simply costs less to have someone muck with.

A similar problem rears its ugly head when it comes time to upgrade. Going from Win95 to Win98 was easy, they were similar enough. Even the 98 to 2000 migration was fairly livable, the programs were the pretty similar. Sure, things like AD caused some heartache, but it was not unlivable.

Longhorn is very different from Windows, and upgrading will be a major pain. Pain in this realm equates to dollars. Add in the scarcity of experienced Longhorn admins, and you have more pain, and more money flow. Not good once again.

Slightly related to people and differences is training. When you move to a new paradigm, even if it was similar on the surface, can cause problems. Again, the 95 to 98 move wasn't bad. 98 to XP with something as simple as a new desktop brought major pain, confusion, and training costs to the secretary set. Longhorn is very different from 98, the problems will be worse yet. Chalk another win up for Win.

All of these issues, admin work, support, training, and ease of upgrades will lessen over time as familiarity with this 'brand new' OS creeps into the enterprise, but that is the future. For now, all the people issues, unless your organization is staffed by IT workers who must be ahead of the curve, will cost you a lot.

The other half of the money pit is software. Not necessarily the cost of the OS itself, but all those other things you need to make it work in an enterprise. This is where the cost starts to add zeros at an alarming rate.

Few software companies have ported their apps to Longhorn, but the number is growing. You can always run your old apps, some of them anyway, on Longhorn as is. Some may require emulator, VMs, or other hacks, but it probably can be done. The problem with this approach is that all the bells and whistles that come with Longhorn are conveniently ignored. Why did we upgrade again if we are just running the same old apps? Good question.

A partial answer for this is the purported greater security of Longhorn. While it has the potential to be more secure, you need a competent admin to lock it down. These are few and far between. While some of it may be doable with a white paper, checklists, and best practices manuals, nothing beats hands on experience. Do you have anyone with years of hands on Longhorn experience on staff right now? Another point in the column for good old Windows, again, Longhorn can't compete on the years of experience front, and would YOU risk you enterprise on something so 'unproven'? I sure wouldn't.

The minor security gains for Longhorn are outweighed by the lack of management software. Most enterprises have a large, rather monolithic management suite to run their computers. If this suite has been ported to Longhorn, you are in luck. Most have not, and probably will not be anytime soon. Software like this is large, unwieldy and complicated, and failure in any way is not looked on lightly by big companies. Porting takes time and effort, so if your package is not on Longhorn, you are probably in for a long wait. Basically, Longhorn enterprise management software is a rare species indeed.

Does it get better for Longhorn? Nope. Custom code must be ported. If you are really lucky, you can just recompile your old apps, and they will work. Again this negates all the advantages of moving to Longhorn, but it is much cheaper. If you have to port, and you almost assuredly will, costs can be shatteringly high to start with, and go up from there. Sitting back and staying with Windows looks better and better now doesn't it?

Then there is integration. Your new Longhorn boxes may not play well with the hardware you have in place. OK, some may work like a charm, Samba and other things that emulate the older protocols will help, but again, you are running the older protocols. If it works, and again, that is a big if, you may encounter weird bugs and assorted headaches. That SAN probably wasn't Longhorn certified when you bought it, right? You can't just slap a sticker on it and make it compatible. If you are lucky, there will be a BIOS upgrade to do the job. Unlucky, and this is much more likely, and you need to rip and replace vast swaths of hardware. This is something no CIO looks at as anything less than a career limiting move. Windows, the safe corporate bet, Longhorn, the not so safe bet.

So, where are you at overall? People, software and hardware are all clear wins for good old Windows over Longhorn. Except for the really rare cases, Longhorn simply costs more. Live with it people. There is a huge, nearly unending support mechanism for Windows, hardware, software and people, all at the beck and call of your checkbook. In the spotty cases where it exists now for Longhorn, it will take a few more pages from that checkbook.

As the MS promoted study rightly points out, few people will do a real TCO study for the cost of migrating from Windows to Longhorn. Less than half if you believe the Yankee numbers. Most will get the proverbial bug up their ass, and jump in. By the time they realize how much Longhorn is costing them it is to late.

If you deploy Longhorn without a good reason, it will cost much more, be a huge strain on your IT staff, and potentially compromise security. General staff will incur slowdowns and productivity loss until training occurs, and even then, won't be nearly as proficient on Longhorn as the old Windows for a long time. If you need help, flip a coin, and if it comes up heads, you may find someone that can help you. If not, where do you turn if there is no one familiar with your OS?

The big picture shows that Longhorn will make you pay, and pay, and pay, and pay. Sticking with good old Microsoft Windows, be it 2000 or XP is the safe choice, you know what to do, who to go to, and what it will cost. Familiarity breeds content in this case. If you are going to move to Longhorn, do two things first. Run the numbers, and I mean really run the numbers, don't just compare the initial sticker prices. Then find someone who can support you to the level that you have already. If a minor miracle occurs, you will be ready to move, if not, don't.

Moral of this story, the arguments put forward by MS are intentionally limited in scope to show itself in a good light. If you look at the big picture, there are some, lets just say 'issues' with their argument. It just goes to show, with 'impartial studies', you get what you pay for. Pay now, and get off the gerbil wheel, or pay later, and stay on it. Your choice. µ

(Note: I wrote the first part with the simple aim of regurgitating the best of MS's arguments against Linux. Once written and revamped, I then did the search and replace. No editing was done once the word Longhorn was grafted in, it stands as it does on merit. I also did not actively try to write the original with the aim of substituting it with Longhorn.)