IBM Reports Indicate Linux TCO Is Lower

Tontoman writes "Information Week reports that two research reports sponsored by IBM argue that Linux is less expensive to buy and operate than Windows or Unix. The first, a Robert Frances Group study, concluded: 'Linux is 40% less expensive than a comparable x86-based Windows server and 54% less than a comparable Sparc-based Solaris server. The Linux server's costs were $40,149, compared with $67,559 for Windows and $86,478 for Solaris.' The second, a Pund-IT report, titled 'Beyond TCO--The Unanticipated Second Stage Benefits Of Linux,' indicates that 'Linux is enormously popular among IT staff members, many of whom are at the beginning of their careers, as well as with IT educators in universities and technical institutions worldwide.' This has resulted in Linux playing a significant role in the recruitment and retention of IT staff and managers."

Comments

TCO vs. HMO

[Anonymous Coward]

IBM cited one reason behind Windows higher TCO: medical bills incurred from employees banging their head on their desk.

a couple of surprises in article

[yagu]

From the article:

it found that Linux is 40% less expensive than a comparable x86-based Windows server and 54% less than a comparable Sparc-based Solaris server. The Linux server's costs were $40,149, compared with $67,559 for Windows and $86,478 for Solaris.

I am not surprised at linux's lower cost, I am surprised Solaris was so high. Other than Sun's high licensing costs I'm at a loss on why Solaris would be so much higher. I've read other studies and I tend to find them credible that one of the biggest cost-savings in TCO is the manageability of a unix-like system vs the Windows GUI approach. I've seen narratives where good unix administrators can sometimes manage at least twice as many systems as good Windows administrators, sometimes more. This is largely because of the simplicity embedded in the unix complexity (one of the biggest complaints I see about unix is its "too-hard" nature, but when mastered my experience has been you can script and automate so many unexpected scenarios easily, something not so readily available in Windows).

The second surprise for me, also from the article:

"second-stage" benefits that some companies are experiencing by implementing Linux. Second-stage benefits expand upon initial benefits such as lower hardware and licensing costs to include the ability to consolidate server workloads, reduce IT hardware upgrade costs, and attract new IT workers interested in open source. The Pund-IT report, titled "Beyond TCO--The Unanticipated Second Stage Benefits Of Linux," indicates that "Linux is enormously popular among IT staff members, many of whom are at the beginning of their careers, as well as with IT educators in universities and technical institutions worldwide."

It's encouraging to note linux is enormously popular among IT staff. Maybe unix and linux have more purchase on the IT world than we thought. I'd resigned my professional life to watching the MS juggernaut conquer the technology world but maybe the unix paradigm has legs! (There are other equally interesting "better" architectures, (Be, Plan 9) but probably are in the wrong place at the wrong time to gain much mindshare.)

(As an aside, have you ever noticed, the admin energies for Windows' environments goes to keeping the system running in as stable a manner as possible, while admin energies for unix's go to extending and enhancing the systems' performance, sometimes in elegantly exotic ways? Just my $.02)

Re:a couple of surprises in article

[PIPBoy3000]

I'm admittedly a Windows person for the most part, as that's the environment I live in at work. The good news about the GUI-based environment is that it's typically fairly easy to pick up a new Windows tool and figure it out. For the semi-casual administrator/developer, that can be immensely useful.

The problem is that after a certain point, it becomes difficult to figure out complex issues. When bugs pop up, it's hard to know whether it's the software's fault or your own, with no good way to peek under the hood. When trying to extend beyond an application's capabilities, you start running into hard-coded issues that make it difficult or impossible.

We're currently migrating to ASP.Net and having internal struggles about whether or not to use Visual Studio, for example. I personally dislike being hampered by the interface, though it makes certain things much easier. The catch is that you need this bulky environment in order to work with what you create, you can't easily edit things outside of the environment, and often the application creates code for you that isn't quite what you want.

So, I'm not sure there's a clear TCO value for these sort of things. Each OS and application probably needs to be evaluated for what you're trying to do. My guess is that there will be a mix of the two systems for a long time into the future. Competition is good.

Re:a couple of surprises in article

[Master of Transhuman]

"The advantages of having inline help, syntax coloring, auto-completion, project organization, etc. in the native home of ASP.NET editing are just too many to think that another environment like a pure text editor would be a good idea."

I don't know too many text editors that DON'T have most of those features, albeit some of them may not as fully support ASP.Net.

The issue then becomes which IDE allows getting under the hood while still providing sufficient automation to enable productivity.

Meanwhile, the main point of the OP's comment was that a GUI (and by extension, closed source) conceals one's lack of direct knowledge of what is going on - knowledge that becomes critical when something goes wrong.

It's constantly true on Windows - something doesn't react the way you expected. On Linux, you can look at a config file. On Windows, you can't look at anything but some checkboxes scattered over half a dozen different dialogues and menu options. The only way to figure anything out is to step up to the next level and reconsider the entire process you're trying to do - essentially relearning the Windows interface for the process every time. Why? Because in fact it's terribly complicated. The GUI just makes it SEEM simple.

I keep telling people this, but they don't listen: Windows is totally NON-intuitive. It's operation is incredibly complicated and deliberately so - first, because it's Microsoft's way to use "featuritus" to lock in its customers, and second, because Microsoft has no clue how to make anything simple.

People think Windows is easy to use because you can point and click to copy a file or something. That's trivial. Try running one of their servers. Try even understanding Active Directory, or Group Policy interactions between the several different types of groups allowed. It's a conceptual nightmare.

Re:a couple of surprises in article

[mchawi]

On Linux you can look at a config file that *gasp* gives you the same information that those checkboxes are - the settings that the system is running under.

Once Linux / Unix / Windows / Any OS has a massive failure - it is complicated to troubleshoot and you need knowledge of how the server and applications work. It's a conceptual nightmare.

In other words - if you talk to a good Windows admin they'll think that the Linux system is a conceptual nightmare because they're used to Windows. If you talk to a *nix admin they'll tell you Windows is a conceptual nightmare because they're used to *nix.

Basically if you don't know the underlying architecture in either system and try and just fake things by guessing - you're not going to get far in a real problem situation. I don't see that as a benefit or drawback of Linux/Windows - just a fact of life. Good administrators have a lot of knowledge about their systems and environment.

Re:a couple of surprises in article

[ArsonSmith]

That makes the best description of windows vs unix usiability I think I've ever heard.

Summed up as:
Windows makes easy things easier and hard things harder. Where as Unix makes hard thigns easier but easy things harder.

Windows low cost of entry expensave maintance, unix high cost of entry, lower maintance.

Interesting Way to Jab at Both Sun and MS

[Metzli]

The article says that the study shows Linux to be cheaper than either Microsoft or Sun. Gee, I wonder why AIX wasn't included as a Unix variant?

see .pdf

[Janek Kozicki]

Actual .pdf of the study here.

Imagine that...

[It doesn't come easy]

But Linux's licensing-cost edge is likely to wane as Microsoft and some Unix vendors, notably Sun Microsystems, lower their prices.

Competition drives prices down...who'd of thought...

A review with numbers!

[rob_squared]

Reviews having actual dollar amounts I tend to trust more. Yes, IBM can be considered biased because, well, they use Linux, and also deal with Microsoft.

What I really want to see, though, is an item-by-item document included for download which shows what they included in their TCO estimate. Statistics and numbers are fine, if you can read the whole dataset for yourself.

How is this news?

[notdanielp]

Per-OS TCO does not exist in a vacuum. Organizational direction, sunk costs from previous IT investments, interoperability with business partners / clients / vendors... each of these is a factor that will be different for EVERY business making the Linux/Windows/etc. choice.

IMO a well-run organization will have a hybrid environment.

That being said, it is useful for planning purposes to know in which situations Linux TCO beats Windows and vice versa.

How do you calculate ?

[karvind]

How are these cost calculations done ?

(a) Maintenance costs

(b) Support and systems administration costs

(c) Application-server support and system administration costs.

Are these really fixed costs ?

Whilst I agree with this...

[m50d]

remember that IBM has a substantial interest in Linux. If it was the other way around we'd be crying foul about how studies will always find in favour of whoever's funding them. Anyone know if there's ever been a truly independent comparison

What about

[Tmack]

X-86 based solaris and Sparc based Linux? While I dont work with any of ther former, I work with many of the latter. While the hardware for sparc costs more than similar X86 hardware, does the TCO for running Linux as opposed to Solaris make up for that extra hardware expense? Does running Solaris on X86 increase the TCO?

Tm

My guess is

[Anonymous Coward]

Other than Sun's high licensing costs I'm at a loss on why Solaris would be so much higher.

My guess would be:

1. Nobody knows how to use it, everybody coming out of school these days is used to using Linux and/or BSD, from this perspective Solaris does a lot of weird things for no reason.
2. Much as Sun's pushing Solaris/x86, if you're using Solaris, you're still pretty much going to be using expensive, locked-in Sun hardware. (Of course that hardware is probably more reliable, but sometimes lower TCO means you get what you pay for).
3. Sun is a competitor to IBM who commissioned the study, maybe the study misrepresents Sun TCO in some way.

Re:My guess is

[j-cloth]

Perhaps, but the argument is now opposed by the second study that says how many eager new IT people have Linux skills. Young = cheap.
This is brining the admin cost of Linux down to the point where Windows admins were a few years ago when everyone got their MCSE.

Re:My guess is

[dajak]

My guess would be:

      1. Nobody knows how to use it, everybody coming out of school these days is used to using Linux and/or BSD, from this perspective Solaris does a lot of weird things for no reason.
      2. Much as Sun's pushing Solaris/x86, if you're using Solaris, you're still pretty much going to be using expensive, locked-in Sun hardware. (Of course that hardware is probably more reliable, but sometimes lower TCO means you get what you pay for).
      3. Sun is a competitor to IBM who commissioned the study, maybe the study misrepresents Sun TCO in some way.

Sparcstations are just too reliable. We have machines from 1991 running NIS+ and some other stuff. No manager making a purchase decision is ever going to believe that a server will run for 15 years without a glitch, and he is not going to spread the TCO over 15 years. Nobody in the organization is qualified to touch the machines, and many of the windows system admins who have taken over don't even know they exist.

The windows admins occasionally screw up the network (like when they made the NIS+ servers unreachable by changing the IP numbers of the only two sparcstations allowed to access them), and then we immediately hire an expensive external admin to solve the problem.

Lessons:
- Sun hardware is too reliable: the machines will be technologically obsolete before they fail. Sun can save costs there, because nobody appreciates it anyway unless they are building a spacecraft or nuclear power plant.
- Comparing an x86 machine against a sparcstation based on a lifespan of 5 years is completely unfair. We spend an expensive two weeks configuring a new sparcstation, and then let it run for 15 years. The Windows machines are tinkered with all the time by cheap Windows idiots. The sparcstation gets cheaper as time progresses (if Windows administrators cannot interfere with its operation).
- What about the costs of letting Windows idiots tinker with your infrastructure all the time? THEY are the ones that create the problems for the sparcstations in our organization because they don't know what they are doing.

Re:My guess is

[Master of Transhuman]

Or one could look at this as a problem:

"Nobody in the organization is qualified to touch the machines, and many of the windows system admins who have taken over don't even know they exist.

The windows admins occasionally screw up the network...and then we immediately hire an expensive external admin to solve the problem."

In other words, you have obsolete machines running critical processes that no one knows how to maintain, so you have to hire external people to solve it.

This is what will happen to Windows or Linux or any other OS if you let "if it ain't broke, don't fix it" rule for too long. (Of course, Windows won't last that long anyway, but that's another issue.)

Just because something works doesn't mean it's not obsolescent. I don't care what it's doing, a fifteen-year-old machine is obsolete NOW.

In other words, it's incompetent management that is the problem, not the OS.

Re:My guess is

[51mon]

"I don't care what it's doing, a fifteen-year-old machine is obsolete NOW."

The hardware may be obselete, but if it is still doing the job you replace it when it fails (or ideally just before). Not having a replacement plan could be an issue, and I suspect these people don't.

The idea there is some perpetual upgrade path we all must walk is a myth created by the IT industry to keep sales figures high, and sustained in part by bad software engineering.

It isn't even obvious they have a management issue, just because they get outside help to sort problems on the boxes, if they only have an issue every few years it is cheaper not to employ the expertise.

I've had 10 year old systems fail whilst still under vendor support contracts, fixed and returned to service inside 24 hours, why should we have replaced them if the economics didn't justify it?

Re:My guess is

[MrSenile]

Working with Sun equipment from the standard Blade 150's up to the Sunfire 25K's, I can say the following with ease:

1) The hardware is DAMN expensive. If you don't have sufficient support, you're paying through the nose for replacement parts. We're talking $3000 for memory. That's three thousand here. For a single DIMM. Gets really nasty when some of the mid-sized servers take 40 such DIMM's.

2) The support for servers is ALSO damn expensive. Talking Platinum service? Get our your wallet. They're 'nice' enough to give 10-30% (and sometimes 50%) discount on hardware/parts to corporations who are tied heavilly into their business, but that still makes support contracts really damn expensive.

3) The support for sun currently SUCKS ROCKS. When platinum support should have a 2 hour turnaround, and we constantly have 24 hour turnaround, that tells you something about how Sun runs their business. I guess it happens when you lay off all your expert personal and hire green off the street college students. No offense, but that's the facts.

4) The requirements are not NEARLY as reliable as it used to be. A sun box before you could shove into a dusty closet and have run for years and years without touching it. However, now, based on contractual support obligations, you're _required_ to keep up to date with patches (even if you don't need them), _required_ to keep up with firmware upgrades (even if you don't need them), and _required_ to keep explorers updated on the boxes (which while isn't a big deal, is a pain in the ass even automated). So instead of keeping up that wonderful uptime of years, you basically have an uptime of a few months, and sun boxes arn't intended to be constantly rebooted. It shows.

5) The hardware is below par. I don't know if they do what the Romulans used to do and get hardware on the lowest bidder, but it's currently sucking. 1 out of 5 systems we get have some type of hardware issue in the first 3 months. Memory, harddrive, case, motherboard, nic, fc-al, something always is going wrong. And frankly it's disgusting. PC hardware is proven more reliable, and linux right now looks damn nice compared to sun.

So you want to know why the TCO for Sun has been so high? Look above, you have your answers.

Bullshit research

[mwvdlee]

I don't need to RTFA to know that IBM is making pretty good money of Linux. No wonder the "research" says Linux is cheaper.
It's the same as Microsoft "research"; 100% pure marketing drivel.

Why do I get the feeling...

[gowen]

that an IBM-funded report favoring Linux won't get treated with the same healthy scepticism that a Microsoft-funded report favoring Window.

Folks : if you treat any of these studies as anything other than another form of advertising, you're a fool.

Salary

[knarfling]

My first thought at why it would be more expensive for Solaris would be that an experienced Solaris Admin can command a much higher salary. Although we would like to believe that a good Linux Admin can work with all kinds of Unix/Linux varients, there are enough differences between them that an admin with home or small business Linux experience might have a little difficulty on a larger Unix system.

It could also be that because Unix is perceived to be a "big busness" operating system, companies are willing to pay more for someone who has actual Unix experience rather than Linux training.

Re:Nice Result, But...

[drnlm]

Is it any more credible than MS studies, no. However, in certain management circles, the MS studies are considered very credible precisley because they're backed by MS.

This study will be very useful as a counterbalance to the MS-funded studies, andgiven that it's backed by IBM, it can't be as easily ignored by management as some of the other, recent refutations of MS's results.

News, no. Good PR, most definately.

WTF?

[tomstdenis]

We built 8 gentoo linux boxes in the span of two weeks here at my office.

Cost of parts: 10K
Cost of labour: two people x two weeks x 900/week = 3600$

Other costs [power/netaccess]: trivial

So for [round up] 15K we bought, assembled, built, installed and setup 8 boxes. that's a cost of roughly 2K each.

Whoopy doo.

Where the hell does 43K/yr come from? Is that the cost of the employee as well?

Well the guy we did hire to manage this, had we planned on keeping him would cost ~60K/yr which when you split over the 32 boxes in the office is trivial.

And we don't have to buy server license upgrades or what not. So really the cost of ownership above the staffer we would have had to have anyways is ZERO. Not 43K/yr.

Tom

So True

[Comatose51]

.Linux is enormously popular among IT staff members, many of whom are at the beginning of their careers, as well as with IT educators in universities and technical institutions worldwide.

That statement is so true. Back in college, we all developed on Linux environment because: 1. Our professors were old school and know Unix and C. 2. More importantly, we can get down into the nuts and bolts of the OS. It really helps when you're taking a class on OS. My friend and I wrote a 2 line Perl script to create and kill process one after another just to see how Linux will handle process IDs wrapping around and basing our design decision on that (part of it is also the Geek factor to see what happens). 3. Linux and open source tools are freely available.

Now at work, most of the younger developers and IT staffers are also Linux users. MS haven't done so well in winning the hearts and minds of the next generation.

proprietary Unix is expensive

[StandardDeviant]

And it's not even the up-front hardware costs that can kill you (Solaris 10 on an opteron is actually pretty damned price-competitive), it is the relative rarity of the applicable skillsets (and there can be a world of difference between a high-end Solaris, AIX, etc. machine and your common linux server on Dell hardware or whatever) which leads to increased salaries for the in-house administrative staff and the cost of vendor maintenance contracts which tend to be much higher than you might expect coming from the windows/x86/etc. world. (On the other hand, with proprietary Unix you do sometimes get what you pay for. High-end support from a single vendor who provided both the hardware and software in a system can be pretty reassuring if you have a business-critical system, and proprietary Unix runs on hardware that in some cases can do things that your common x86 stuff just does not scale to, both in terms of reliability and in terms of capability. As with all things, tools have jobs they are better suited to than others.)

How does this change anything?

[TheCabal]

Every time MS puts out a report that Windows TCO is lower, everyone here dismisses it as propaganda. What about this time? IBM has a substantial investment in Linux and I noticed that their own AIX wasn't used as an example. It's just another case of manipulating the facts to fit one particular view. To call it anything else is intellectually dishonest.

That deserves an +5,Insightful

[Lonewolf666]

The good news about the GUI-based environment is that it's typically fairly easy to pick up a new Windows tool and figure it out. For the semi-casual administrator/developer, that can be immensely useful.

The problem is that after a certain point, it becomes difficult to figure out complex issues. When bugs pop up, it's hard to know whether it's the software's fault or your own, with no good way to peek under the hood.

Exactly my experience, and I'd like to add that Microsoft online help tends to be similar:
Basic tasks are well explained, but once you need help with complex issues, the approach of "open this window and click that button" breaks down. At this point you need information about how the application works, and that is usually absent in the help files. If you are lucky, you can find it online in the MSDN, but even that tends towards pre-formulated solutions.

Slashdot biases

[allanc]

You know, I'm seeing a lot of posts saying things along the lines of "If this were a Microsoft study, everyone would be calling foul, but since it's IBM and it's pro-Linux, everyone's going to accept it unquestioningly!"

I have not actually seen any posts accepting it unquestioningly. At least none getting significantly modded up.

So, you know. Calm down. Talk about the actual article, don't just complain about Slashdot.

(Yes, I know this post is hypocritcal)

Don't believe either of them

[jim_v2000]

And use whichever one you feel most comfortable with, because in the long run, the cost of having the box sitting there is going to be about the same. I'd bet that most of the cost difference just depends on the IT staff. I'm sure that there are experienced Windows and Linux IT guys that can keep their respective boxes running well for little cost. It's when you get bumbletards running around trying to be IT that causes the TCO to rise.

Re:but but

[MightyMartian]

Torvald's Heroes

Colonel Gates: No vone escapes from Stalag XP!

Torvald: Hah! We have a far lower TCO. We don't need your virus-laden operating system.

Colonel Gates: Tell him, Ballmer!

Major Ballmer: I know nothink!

Colonel Gates: Torvald! Nothing can stop Vindows now! Ve have unstoppable software!

Torvald: You'll have to hold on a second, I think Major Ballmer thinks your desk is apple strudel.

Colonel Gates: Relax, Torvald, Major Ballmer is simply practicing for ze next trade show. He's hoping to injest ze vile Steve Jobs. NOw, back to your Linux. It is bad, and smelly, and costly, and is made by Communists!

It's funny...

[.com b4 .storm]

It's funny how Microsoft funds a TCO study that shows Windows to be lower in cost, and the Slashbots rise up and flood the comments with "well of COURSE that's what a MICROSOFT funded study will show." Yet when IBM does the same thing, there is a distinct lack of comments of the same sort. Newsflash: corporation funds a study and the results miraculously serve its interests!