Linux TCO: Less Than Half The Cost of Windows 700
ggruschow writes "Linux Today reports 'The cost of running Linux is roughly 40% that of Microsoft Windows, and only 14% that of Sun Microsystem's Solaris, according to a new study which examined the actual costs of running various operating systems over three years.'"
Microsoft says so, too! (Score:5, Informative)
Re:Solid Unbiased Reporting... (Score:5, Informative)
Don't think they work for Linux Today, unless Linux Today has 14 Global 2000 Enterprise businesses.
DanH
Gross pay is less than half the cost of employment (Score:5, Informative)
So the majority of the costs are based on the server admin's cost, which averages $71,xxx a year...my question is, where are this jobs as linux admins for $71k/ year?
An employee's gross pay is typically less than half of what it costs to employ him or her. An employer needs to buy office space, power, lighting, air circulation, health benefits, not to mention the employer's share of the taxes (in the USA, payroll tax and Social Insecurity matching payments).
Re:Microsoft says so, too! (Score:3, Informative)
when he says "have a high price" he is talking about initial cost compared to linux
Now whether you believe that or not is a different thing, but his quote wasn't contradictory.
Re:first? (Score:3, Informative)
Re:Microsoft says so, too! (Score:4, Informative)
Re:first? (Score:4, Informative)
Unix/Linux TCO book (Score:3, Informative)
You can download some parts of the book for free to get a flavour of what it's about. I actually bought a copy and would recommend it for anyone thinking about converting from Windows to Linux - it's only $30.
Re:first? (Score:5, Informative)
Apparently they were calculating the TCO of webservers running on Linux vs webservers running on Windows and Solaris.
This had nothing that I could see about running app servers, file servers, databases, etc...
Not to mention that the Windows installations used IIS and the *nix's used Apache. So it doesn't answer the question: What if Windows used Apache? Which may reveal a slightly different result (and would show that the measurement are actually about IIS vs. Apache, not Linux vs Windows).
I'm not saying that their data or methods are crap, just that like with any stats, be careful to read the sources and methods behind collecting and collating the data. Look behind the presentation.
Re:Yea But.. (Score:2, Informative)
Start getting into application servers, that's another story, but really, there's nothing you can do on Windows/IIS/Activex that cannot be done on Solaris/Iplanet/Java including (Centralized Authentication, Database Access, Server Side Dynamic Code, Client Side Dynamic Code, etc.)
Only difference is it costs 1 admin to maintain 10 NT/IIS webservers and 1 admin to maintain 30 Solaris Servers. Now add scaling onto that where one Solaris Server can hold 2 or 3 times as many processes/applications than an NT box. Also add that it is cheaper to reach 3/4/or even 5 9's with Solaris/Iplanet, and you get these Total Cost of Ownerships.
Finally, I haven't read the artical yet, but am very familiar with webhosting issues, it is imperative to compare level of support. If you are looking for 99% uptime, I bet the numbers between OS's are very similar, and probably Solaris and even Linux looks expensive. However, aim for 99.999% uptime, and you will realize that NT/IIS just isn't capable of reaching that level. To reach 5 9's you almost have to go with a real Unix on real (non-intel) hardware.
To reach 5 9's on an NT environment, you will end up spending between 2 and 10 times more than with Solaris. (That's because after fighting with NT(4.0 and 5.0) for a year, you'll eventually sell the hardware and licenses and build a Solaris environment)
Re:Consider the source (Score:5, Informative)
Meanwhile, here on /., people seem to be saying that the report came from Linux Today, and therefore is too biased to be trusted.
So on the one hand, you're wrong about the source of the report, just like a lot of other posters. On the other hand, you're wrong about /.'s response to the report.
But hey, at least things are somewhat better than you expected, which is always pleasant.
Warm Body Effect (Score:4, Informative)
The reason for this is the warm body effect. A single Windows machine is actually easier to manage than a typical Linux or Unix (AIX in my case - I haven't used Solaris) box. However, this is due to lots of GUI tools.
The problem with GUI tools is that they require a warm body - you have to physically sit in front of a console and type on the keyboard or move the mouse. While Windows does have scripting support which you can enable (and install decent script languages such as Python), most management tools don't provide a scriptable interface. If you are lucky, they have an API, but writing a script wrapper for a programatic API is a lot more work than simply invoking a command line interface.
So as long as you are sitting in front of the machine, Windows shines. If there are lots of boxes in the room, Windows in a pain. If there are lots of remote boxes, Windows is a pain in the rear.
*nix boxes, on the other hand, while frustrating on the first box while you try to locate various config files in random locations, are a true joy to manage as you expand your control with custom scripts to work for and report to you - thousands of daemons like a magician of old.
Perhaps... (Score:4, Informative)
Re:The study (Score:4, Informative)
However, I agree that training costs are part of the TCO. It is raised everytime I suggest a move to Linux. I then point that while there is a cost in retraining when moving from MS to *nix, it is pretty much a one time cost. Unlike MS systems where you get a new (sic) system every 2 to 4 years. Immediate training cost may be larger but the ongoing training costs are much lower.
As for the printing question, in my experience, the cost of printing problems plummets in a *nix network -- primarily because the printer servers don't fall over when the moon is in the third pahose or someone looks crosseyed or . . .
Re:The Cost of Downtime (Score:2, Informative)
Re:GUI bad, CLI good? (Score:3, Informative)
While true you sometimes have to buy or download these tools, this doesn't mean that you can't use em. It's been a while since I messed with this stuff, but when I worked in Nix/NT operations I remember some of the command line stuff on NT and in the reskit was ok. Also there's some great stuff out there like Roth's Admin Misc perl modules which are just great (see http://www.roth.net/perl/adminmisc/)
Re:first? (Score:2, Informative)
You'll probably get modded down for that, but I would agree with you. You have to know and have used the product to pass most of the tests, especially now in the 2000 environment. A lot of the questions involve not multiple choice, but you are trying to do task A, here is a mock up, do task A.
I've worked with some MCSEs that I wouldn't trust to run an NT server, but I also am an MCSE and know a lot about running NT (I've only just started on my 2000 cert, mainly because I'm just starting to work with 2000 in a server environment).
Re:Windows XP for free...if you are a student (Score:3, Informative)
Umm... maybe this is true for you. But let's face it, MS only ever gives away product as a prelude to 2 events: 1) bundling the product into the OS or Office 2) increasing market share via free product then charging that newly gained market share on the next go-around.
Do you honestly think you're going to get Windows for free once you're in the public work force?
On the plus side, Windows has become MUCH to operate in server environments and even on the desktop but, if the study is to be believed, it still has some room for improvement.
Number security advisories this year (Score:2, Informative)
Red Hat 77 [net-security.org]
Microsoft 70 [net-security.org]
FreeBSD 25 [net-security.org]
Sun 6 [net-security.org]
Novell 4 [net-security.org]
Re:first? (Score:2, Informative)
Try Terminal Services and Windows Scripting Host. I am not a Windows SysAdmin, but I know many. The last company I worked for had one SysAdmin for about 50 Win2K desktops, ~12 Windows servers, 1 Linux box, and two unix-based network storage devices. The servers were all offsite aside from a couple of test machines. Using Terminal Services he could easily bring up the remote machines to configure them. Using WSH, he scripted IIS config's, user management, Exchange configs, etc. To add a new web site to IIS or to create a new user one doesn't even need a GUI. I'll be the first to admit that MS should not have GUI's on it's enterprise servers, but it's also not required. Everything can be done at the CLI and script level. This is where your true efficiency is. Now add the fact that almost every configuration in Windows
Comment removed (Score:3, Informative)
Re:Obvious -- except for what you are forgetting (Score:3, Informative)
Lots of people complain about the new licensing. But let's not forget that nobody's got a gun to your head. If you want, you can just pay full price for every computer in your company and be done with Microsoft forever.
Re:first? (Score:5, Informative)
The OS is installed and configured automagically via scripts, and each machine can be completely reformatted/reinstalled by pressing "N" as it reboots.
The back-end is NetWare, with ZEN for application distribution. So no, it's not all Windows, primarily because the university has been pretty much in bed with NetWare for the last 10-15 years.
So when you say that "complex things are unscriptable," that leads me to believe that you have no clue what you're talking about. How is it that editing text configuration files is so much easier than editing text registry patches?
And no, this isn't a "Windows is better!" debate. I just think that if you have people running your systems who aren't morons or zealots, you can make just about anything work well.
--Jeremy
Incorrect information in the report? (Score:2, Informative)
Looking at the hardware costs for Solaris, they mention the "most common servers in use were the Sunfire 4800 and 6800". This doesn't sound accurate and I'll explain why.
Both of these servers are enterprise-class machines usually designed for high-availability, processing-intensive applications, such as databases. Very few sites would use such big iron to run webservers on. For webservers, most companies would use lower-end hardware, such as the Sunfire 280r or the Ultra Enterprise 420r. At a former dotcom I worked at, we had about 400 third-party E450s and 600 Ultra 10s, which were used both in production and development, and were handling an average of 12 million connections every day. We had 8 Ultra Enterprise 6000s (the precursor to the 6500 and 6800) which were exclusively for databases.
In addition to this overkill in hardware costs, the report is also incorrect in the licensing costs. They mention the vast majority of customers used 8-CPU systems (very strange; the 4800 can hold 12 CPUs and the 6800 up to 24; it'd only be sensible to max out the boxes to offer better performance), which makes the licensing costs claim dubious at best. The cost of a Solaris RTU (Right To Use)license upgrade for a 5-8 CPU box is $6,000.00, not $12,500.00, as listed in the Sun Store.
So, either the report's data is slightly askewed, or they picked the wrong companies to use as models for this study.
Just my two cents...
Re:first? (Score:1, Informative)
My class was given by a CompUSA instructor in South Florida. Most of the examples and questions were similar to, "What Microsoft solution should you sell ^H^H^H^H^H propose to the customer?" I.e., it trained students on how to sell Microsoft products. Now I'd agree that very few could actually take this test cold and pass; the tests do cover a broad range of *M$* products. However, these cram centers really know what will be asked and really gear their training towards passing the test rather than teaching sound networking concepts or troubleshooting skills.
Of the ten or so MCSEs at the company where I worked at the time, fewer than three could explain sub-netting adequately. All knew the OSI model, but only a couple could explain what each actually did.
Figures are WAY out (Score:3, Informative)
They equate single and dual-CPU commodity x86 boxes with 24-CPU US3 servers with 100% redundancy for guaranteed uptime.
No wonder their figures are utterly bogus.
If you take their own calculations, factor in COMPARABLE figures across all 3 platofrms, then you get Windows as the most expensive, Linux second and Solaris woith the lowest TCO.
But then, that would not have made for a good story, would it...