Ask OSDL CEO Stu Cohen About Linux TCO Studies

This morning OSDL and OSDL member Levanta jointly released a study done by Enterprise Management Associates called Get the Truth on Linux Management. For years, a proprietary software company in Washington State has run what they call a Get the Facts campaign about Linux, full of studies that invariably show Linux to be expensive, hard to maintain, and less than totally secure. Stu Cohen, as CEO of OSDL, a group "dedicated to accelerating the growth and adoption of Linux in the enterprise," will happily answer your questions about Linux vs. Windows studies and the myths and FUD that seem to hover over them. Expect Stu's answers to the 10 - 12 highest-moderated questions later this week.

A Movement within the Students

This may seem like an inane question but why don't I see more of a push to get Linux into the realm of academia?

I know that Ubuntu [ubuntufund.org] has made strides to incorporate themselves into learning environments but where is the effort to alert students (primarily other than computer science majors) to the benefits of Linux?

When I was a freshman at the University of Minnesota, a friend handed me a CD distribution of Debian that would change my life. I knew of the Linux labs in the University but only now did they interest me. I'm now getting my masters at George Mason University and I don't believe there's a single Linux machine on campus. In fact, the whole Computer Science department has only two Sun servers to offer me an account on! Everything else is Microsoft!

Now you may lay claim that every computer science major these days is running Linux anyway. But how about the other areas of study? I used to take music theory and people would rant and rave about their Macs or one of various composing suites in Windows. I tried explaining that Linux has (certainly more affordable) solutions to offer in this department too but no one would even listen to me. It's not like they were mixing platinum selling records, they were just looking for software to write sheet music with.

I think that both Apple and Microsoft realize that the toys people have in college become the toys they demand in real life. So there are all these [apple.com] efforts [e-academy.com] to garner the student's interest hoping that they will use them in their careers.

They make it free (which Linux already is), they make it easy and they make it available.

So how about it? Why isn't the Linux community minting install discs and distributing literature on campuses? Why isn't Linux tailoring cheap solutions to K-12 schools that don't have the money for Windows anyway? Why do we risk letting someone leave academia without ever experiencing the real fruits of it?

If you are doing this (and I just don't know about it), what steps have you taken?

Is it about Linux or better operating systems?

What I would really like to know is why Linux or Windows? Why hasn't there been a really good study that included BSD, Solaris, OSX, or even licensed variants of Unix? Is it all about Linux or is it about better operating systems?

This doesn't make sense

Why would you expect that the answers of someone "dedicated to accelerating the growth and adoption of Linux in the enterprise" would be more objective, in any way, than any of the reports created by pro-MS companies?

It just doesn't make sense...

Bias

Since almost all of these studies are funded or organised by a party which appears to be inherently for or against one of the things being studied, will it be possible to find anyone willing to compare them impartially? After all, how many people would believe an Open Source company to be any less biased than MS when it comes to comparing their products?

Counter spin .. :(

So basically this Q&A session is just spin in the opposite direction to the Windows spin?
Are there any really independent studies on TOC that are produced by fanbois of one side or the other?

Security Question

How can we fix the problem of the way TCO studies handle security? In so many of them every OSS application under the sun gets tallied against Linux systems, regardless of how obscure, or unrequired that application may be. Yet all of the 3rd party things that have holes in them rarely seem to even get looked at when talking about Windows security. Firefox for example seems to get tagged frequently when talking about Linux security in these studies, but Firefox isn't integreated into Linux, and it runs on both platforms. IE on the other hand is integrated into the OS, sure you can not use it, but there is a ton of junk in Windows itself that requires the various bits and pieces of IE to operate correctly. What is it going to take for these studies to finally start comparing apples to apples in regards as to what really is part of the OS and what is required for it to run?

I Wonder What

The Total Cost of Owning Enterprise Management Associates is? Willing to bet the folks behind the study know...

One question

How many rounds would you go, one on one, against Steve Ballmer in an auditorium full of chairs?

One of the main problems

In looking at Microsoft's TCO claims in particular, I've been unable to avoid noticing that a lot of the company's material on this subject consists of, to put it simply, straight lies. Aside from anything else, nothing is mentioned by them about their licensing fees. How they can state with a straight face that after their licensing fees, Windows can still be cheaper than Linux is beyond me.

Legitimate performance competition is one thing, but I'm curious to know how the ODSL is able to deal with Microsoft's lack of ethics in this regard? Given Microsoft's marketing power, how are Linux advocates able to communicate to people that many of Microsoft's claims in this area are deceptive?

Setting up Linux from Win2K3

Say I wanted to switch from Windows Server 2003 to Linux in a company of about 400 people with the same equipment I already have, generally speaking how long would it take and how much would I need to invest?
Do I need to hire several Linux experts just to get it up and running?
Would you expect this to be relatively easy or would it be very complicated and time consuming?

To be fair

This study is sponsored by the OSDL so it has an initial bias. If it were a study proving Windows to be cheaper sponsored by Microsoft, everyone would be yelling at the bias - rightfully. So it's only fair game to strongly underline this. (BTW, I AM Linux biaised. but that's not the point here ;) )

Which is better? It all depends!

Especially not in a heated market like the OS biz. Who can tell what's "better" or "worse"? To what scale do you measure? And even if you find a way to compare them, what tells you that we won't see the same phenomenon that benchmarks sparked in the CPU and Graphics sector, companies that trim their products to perform perfectly in the artificial test environment (and really suck sometimes in everyday appliances)?

Do I need graphics on a server OS? Do I need highly sophisticated user permissions on a single user machine? Do I need support for 10 billion hardware pieces? Do I need flying pages when copying? Is it important that you can trim the system to run even on a P90? Do I want to be able to use the most recent fads in anti-aliasing and pixel shading? Do I need to be compatible with 100 other formats across 20 OSs? Do I need or want to customize my kernel? Does it make sense to cram the GUI into the system (and the internet browser as well)? Is it useful to ram the Mailreader into the system so tightly that it's virtually impossible to get rid of it?

No offense, but who are you to answer those questions for me?

So which system is "better"? Neither. Or both. Or it's really one of them. It just depends on who you are, how much you know (or want to know), how flexible you would like to be, and most of all, what you want to do with your machine.

If OSDL believes that Linux has a superior TCO ...

... why don't they use it?

Almost every PDF document on the OSDL website has been created on a Windows PC or on a Mac. Even the Desktop Linux Survey Report [osdl.org] shows:

$ pdfinfo DTL_Survey_Report_Nov2005.pdf
Title: Microsoft Word DTL_Survey_Report_v4.doc
Creator: Word
Producer: Mac OS X 10.4.3 Quartz PDFContext

Are the OSS IP Indemnification offerings worthy?

We recently had an issue in which Microsoft Office included unlicensed IP (according to a court settlement). Microsoft did not require us to patch existing installations, rather simply protecting our use via the settlement, agreeing to require future installations to include the patch. This seems like a case in which indemnifications worked (although they could have offered some compensation for the extra work - it's cheaper than litigation). For background, see http://www.microsoft.com/windowsserversystem/facts /topics/ipi.mspx [microsoft.com].

How do the OSS indemnification plans stack up? Have there been any significant cases involving IP indemnification?

Linux devices problem

I believe that if you are OK with the application space that you can have equivalent or lower enterprise TCO on desktop Linux.

However, in broadcast engineering, we have a problem that there are lots of devices (satellite receivers, video compressors, video effects devices, video monitoring systems) that are using GNU/Linux. Each vendor seems to pick a different distribution version, basically requiring keeping track of patching 10 or 20 different OS versions. And the truth is that vendors seem so sold on the notion of Linux security, that they often don't feel the need to have to even consider the need for automatic and regular patching of the OS. While Linux does tend to have fewer security problems than Windows, they do come along every now and then.

By insisting on Windows in devices, one can at least know there is a single location for automatic patching. You do have to be on top of the situation and be wary of zero-day events, but it is fairly manageable.

Why Should We Care?

It's a Serious Question. Don't TCO costs end up coming down to how much you will pay employees, how many employees you need, and the price of software? Shouldn't any capable manager be able to estimate the costs themselves? After all, I'm certain TCO varies wildly from workplace to workplace, considering what kind of system is already in place, what software is readily available for an OS, and what skills your current employees have.

My question is: is there really a use for these reports other than for 'defense': positive propaganda versus negative propaganda?

As an aside, do these studies take into account the availability and flexibility of currently extant software? Is there even a way to turn that information into TCO?

Quality comes with price?

Maybe the TCO summaries are right after all?
Atleast partially that is.

Using the linux road, you have to pay competent people salary for actually knowing something about the system they're dealing with.

Anyone can get windows server up and running after 10 minutes of reading help files, but it won't be secure by a long shot.

I guess same applies for linux in some ways, but it's like comparing iron ball and snow ball in hell.
Both will melt down eventually if left unmaintained, but it's just matter of how long it takes.
And longer it takes, the more profit you make.

TCO might be higher, but you simply get more work done when your IT department doesn't have to spend 2 days every week reinstalling all workstations.
And getting more work done increases profits and in the long run, brings down the TCO, even if it's higher at the beginning.

TCO surveys are statistics, and statistics always tell what the collector wants them to say.
It's just matter how you count things.

TCO Claims

Ernie Ball goes Linux [com.com] if you havn't seen it yet. There is alot of noise about these mythical enviroments that are pro windows or pro linux, but here is a good example of a real world switch. Ernie Ball makes guitar strings, so there really isn't any internal bias about who to support beyond it being a business decision. It is also a bit of an entertaining story on how they dealt with the MS strongarming about their licenses.

Linux management study versus Linux desktop survey

What does have more impact on the success of Linux, the "Linux management study" or the "Linux desktop survey"? Which of these two areas are more important and should be taken more care of?

O. Wyss

What difference can OSDL make?

There's certainly been a good few questions asked already, but the one I'd like to get an answer to is,
how do companies see OSDL? Do they believe it's a trustworthy group that knows what they're talking about, or does it look like another one of those 'fad-like' groups that's going to fade away? I don't mean to say OSDL is fading out, I'm curious to know what the real-world perception of it is. I've noticed that while many of my friends use linux and are generally well-versed in what's going on, they're usually totally unaware of the existence of OSDL, or it's purpose.
How will this change? How will OSDL become a trusted group for IT managers, especially in a world where most of them have only heard of Microsoft's "Get the facts", or have some shares in MS stocks?
I feel that part of the reason that one of the above posters was asking why isn't linux penetrating the educational market is because the trustees funding the schools have a say in what to use, because they're paying for it, and the trustees will usually have a significant amount of MS stocks.
What's the chance of all of this changing? Or rather, what are the means in place for all that to change?

It is less than totally secure.

...full of studies that invariably show Linux to be expensive, hard to maintain, and less than totally secure

I suspect the first two are potentially true, but that would depend entirely on the situation. Bad choices can always be made, regardless of the systems involved, that turn out to be expensive and costly to maintain. Just because it's open source doesn't make it immune to bad management.

The third is most definately true. As far as I know there is no OS that is totally secure. It's a lauable goal to be sure, but not one that I ever expect to be reached.

OS Deathmatch

I'd guess the only fair way to pit one platform against another would be to offer a scenario - client company X has a list of specific needs and requirements - and let teams of experts of either party deploy their solution. Mano an mano.

Then when the smoke has settled, they are compared with regards to cost for things such as licenses, staff, etc.

It would also be important to note the differences in the solutions to the client.

Will the MS solution be simpler to manage, to update? Will the Linux solution require less tweaking a year later? Will there be hacks beknownst only to the people who set up the solution.

And to make it all worth while - these contests should be arranged regularly and have different levels of difficulty and scope.

Call it "OS Deathmatch" or something silly like that and offer prices. Host it at sports arenas. Set up a fair with computer gear for sale at the entrance.

Invite thousands of low- and high-profile geeks. Invite crackers to attempt to find vulnerabilities with the solutions.

Invite companies with real-world cases to get the contestants to work on their requirements. Let them sponsor the show and in return get the elite solutions.

Not only would this generate tremendous media coverage and potential income for entrepreneurs, it will also make for much more fair scrutinizing of the software than the current crop of shady "independant experts".

Leap to desktops?

There is a large shout for expanding the amount of desktops running Linux. While most users on Slashdot seem fine and dandy with the way Linux desktop is now, I believe that a lot of changes will have to occur before you can get Joe Sixpack to replace Windows or Mac with Linux, such as making tasks more automatic, improving hardware support, and completely removing the need for the command line/terminal (except for development).

Do you believe that the desktop needs to change before its user base expands? If so, what changes do you believe are necessary, and which would be mere "bonuses"?

The difference to a worker is?

This is really telling. Part of the argument is that linux costs more because you have to pay your staff more. As a tech or someone planing on going to school to learn one or the other which would your choose? I'd rather make more and work with Linux but that's just my opinion. There must be some benefit to working in a windoze shop other than the low pay.

Biggest issue is conversion costs

In rare cases, servers can be deployed using just about any server operating system (BSD, Linux or Windows). Usually, though, history matters. Applications that originated on a Unix type OS are generally going to be much more easily ported to Linux than applications built on MS Exchange, ActiveX controls and VBA. Similarly, conversion costs from Unix to Windows can be very high.

There is also the issue of staff retraining. I am aware that the study looked at availability and costs of Linux versus Windows admins, as well as how much training was needed for existing staff. But, this is overly simplistic if considering massive OS conversions. As a practical matter, you do not want to layoff your existing staff (who understand your entire setup intimately) to replace them with people who happen to have better knowledge of Linux. In some countries, you would not even be legally permitted to do so.

Am I right in guessing that the mix of operating systems in almost all these sites evolved gradually, and the decision for individual servers rarely depended on the kind of TCO evaluation favoured by studies like this latest one?

How to include virus aftermath in TCO?

The biggest risk to running a MS shop is the proven history of exploits on the platform that can wreck havoc on your network. True, if the patches are up to date your risk is greatly reduced, but we have seen plenty of organizations that ought to have been better prepared get clobbered. It is a real risk, and can be as a result of intentionally not being up to date (because the patch hadn't finished QA), or unintentionally (mistake or oversight by the sysadmins).

The problem from a TCO point of view: How do you quantify this in terms of a cost? Many (most?) companies are never affected at all, and thus have no cost. Others have their entire business grind to a halt for a day or two, at what must be a massive cost. So there is most definitely a cost here that should be included in the study, but any figure is almost certainly going to be criticized as either much too high or much too low. How are you planning to address this?

The problem with studies

Studies are usually designed to prove a point rather than to objectively study a difference. Even when this is not the intent, study designers are likely to subconciously inject their own biases. Only a truly representitive sample, compared totally objectively, can really be considered neutral - and you're simply not going to get that in any study conducted by people. There is also the problem that TCO is not a fixed number. You're looking at a fairly complex function that varies wildly according to a large number of parameters. however, any such plot of TCO would be completely meaningless to any corporate manager.

I guess my question could be phrased as: "how do you maximize the usefulness to the audience while minimizing the sacrifices to accuracy? And how much of a compromise can you really afford to make before the study actually degrades understanding?"

nice graphs

Wow - there are some nice graphs in that pdf! I love those pie charts! Thank God for those 'sophisticated management tools'. If it wasnt for them, we'd be screwed!

Vista's impact

Do you expect Microsoft's upcoming Windows Vista release to help or hinder adoption of linux on the desktop and how can the open source and linux communities best take advantage of the sometimes-artificial commercial software upgrade cycle?

Why Windows is sometimes the better answer.

What would you say to the fact that if someone has a problem with a Microsoft product, they can go straight to the source. To the people who really made the product. The way Linux is developed, there are at least two tiers of manufacturing. The lower tier is the Linus and company who are developing the kerel and the high level are the actual distributors (Red Hat, Novell, etc.). In today's world, if you are in charge of a multi-million dollar project of setting up servers/datacenters, don't you want a corporate entity that is completely responsible for their product as a scapegoat?

Hetrogenous Environment.

I have a sneaky suspicion that a sensibly designed hetrogenous computing environment would have the lowest TCO of all. Sadly, since these studies are inherently oppositional, they'll be more likely to polarise computer users into chosing one or the other. What sort of studies would measure and encourage interoperability instead?

Downtime

In every Windows vs. Linux/Unix/*nix {insert your *nix variant of choice here) I never, ever see this addressed:

Microsoft redefines "downtime" for their studies. Downtime, in their terms, does not include "scheduled maintenance windows" during which a system needs to be taken down for a backup, integrity checks, defrags (e.g., the Exchange info store), or reboot for configuration changes or patch installation. Where you might have to reboot a *nix box once a year for kernel patches (on rare occasion more often), it's not like the weekly or biweekly frequency of Windows patches which more often than not require a reboot, which can lead to downtime of a critical system service such as your DBMS, mail server, DNS server, or an active directory global catalog server.

Why do no studies make mention of this, and at minimum compare the "scheduled maintenance windows" and how the REAL downtime is affected? Windows, if PROPERLY maintained to avoid future disasters, will be nowhere near five nines' worth of downtime. No way, no how, unless it's on a LAN where you KNOW access is controlled, patches are unnecessary, and you run an intranet mail server which allows for live maintenance.

Better yet, let's have a study which ignores Microsoft's redefinition of "downtime" and instead uses what the rest of the real world regards as "downtime" - let's compare THOSE numbers.

Don't get me wrong; Microsoft produces some damn fine products and the Exchange/Outlook pair is a perfect example of that - it answers a business need that only recently has been addressed in the *nix world with Scalix, OpenExchange, and Zybil (which I only discovered when reading the latest Linux Journal). Sure there have been so-called groupware solution such as Lotus Notes for a while, but anyone who has used both Exchange and Notes will tell you that Notes sucks. It's bloated, slow, and downright painful, even if your database is well-designed; and not only that, it IS proprietary.

So you are the local PC dealer?

One question in return:
Do you still make a profit on the $469 computer after throwing in 2 hours of lessons by your trained staff?

Re:Symbiotic Sysadmins?

Maybe in 1995. This is 2006. Welcome to the new millenium. The most terrifying my upgrade procedures get is forgetting to type sudo before apt-get.