So, what to do with this. If we don’t run it, we’re pro-Linux. If we do run it, we’re pro-Microsoft. And I’m sure that whatever we do, we’re anti-Apple somehow. In any case, here we go: the latest market share figures from IDC about servers show that Windows is by far the most popular server operating system in terms of unit sales, increasing its market share even further. Linux, on the other hand, saw its market share in the server market sink a little.
I don’t really know what to make of these figures, but alas, let’s just get them out there. In the first quarter of 2010, Windows had a 75.3% market share in terms of units sold, up from 73.9% in the fourth quarter of 2009. Linux dropped slightly from 21.8% in the fourth quarter of 2009 to 20.8% in the first quarter of 2010. UNIX dropped from 4.4% to 3.6%.
When you look at the revenue side of things, the situation is even better for Windows, and also quite good for Linux.
- Microsoft Windows server demand was positively impacted by the accelerating x86 server market, as hardware revenue increased 33.6% and unit shipments increased 28.3% year over year. Quarterly revenue of $5.1 billion for Windows servers represented 48.9% of overall quarterly factory revenue. This is the highest percentage of server hardware revenue that Windows servers have ever represented.
- Linux server demand also improved sharply in 1Q10, with revenue growing 20.4% to $1.7 billion when compared with the first quarter of 2009. Linux servers now represent 16.2% of all server revenue, up 2.1 points over 1Q09.
Architecture-wise, non-x86 systems are taking heavy hits. “The market for non-x86 servers, including servers based on RISC, EPIC, and CISC processors, declined 25.9% year over year to $3.6 billion in 1Q10,” IDC writes, “This is the fourth consecutive quarter that non-x86 servers have been outperformed in the market by x86 servers.”
There are a few qualifying notes here. The big one is this: these figures only account for servers sold with software pre-installed, and completely ignore servers where software has been installed manually. If there was some way to include those as well, I’m sure Linux would see a big boost.
What these figures do show is that despite the usual deriding remarks people make about Windows, Microsoft’s products are simply doing very, very well in the server space. Windows Server 2003, 2008, and 2008 R2 have all been received very well, and it seems like sales figures confirm this.
Seriously though, it’s easy for MS to sell more units than Linux since you don’t need to buy Linux to put it on a server.
True, I was about to say the same. It’s very difficult to acquire any usage statistics for Linux since the latter does not require any product activation or it reports back to any company servers. I have a software project myself which installs on Linux, Windows and MacOS. While it is very easy to track the number of installations on Windows and MacOS by just counting the number of downloads, Linux users use our software by either checking out the code or by using our Launchpad repositories. In both cases, I have no means of telling how many installations we have on Linux.
When Microsoft counts their sold copies of Windows, they also count the OEM versions that ship with computers which people bought to install Linux on. Officially, these computers run Windows.
And, PS:
http://www.top500.org/stats/list/35/osfam
=P
Adrian
When Microsoft counts their sold copies of Windows, they also count the OEM versions that ship with computers which people bought to install Linux on. Officially, these computers run Windows
There are times when the geek needs to keep a stronger grip on reality.
This is one of them.
25% of Steam users are running 64-bit Windows 7.
64 bit has become the default OEM Windows install on everything but the netbook.
W3Schools posts webstats for Linux dating back to March 2003:
2.3%-4.5% share in seven years.
The Win 7 Beta had a 2% share in July 2009. Win 7 a 19% share in May 2010.
The Net Applications webstats are less charitable to both Linux and Win 7. But Win 7 is on track to break 20% this month or next.
While Linux remains stalled at 1%.
Well, that’s not the point. The truth is that all those statistics you report here can be forged by the click of a button. I’m using Google Chrome on Linux here. If I ever stumble across a website which is “optimized” for Internet Explorer, I just toggle the browser agent that Google Chrome reports to the web server and I’m officially a Windows user.
Objection, your honor!
Two things:
1) Steam is to-date available for Windows and MacOS only while the latter has just recently been released; no-one would ever expect anything but high usage shares for Windows in this case. There are rumors for a Linux version and actually working code from Valve which runs on Linux. Once this version is released and has been on the market for 1-2 years, we may come back to this discussion and look at the numbers again.
2) If I run Steam on Linux on Wine, which actually works quite nicely, Steam will also register my user as a Windows user.
So, did I ever contradict to that? The point is that one thing is what many many users do first, is deleting the Windows partition and installing Linux. It’s probably not that common over there in the United States but in Europe people love choice and they actually take advantage of that.
Again, as I said before. Never trust any statistics that you didn’t “tune” yourself. There is absolutely no reliable way of telling how many Linux installations exist world-wide, none. All these comparisons and statistics are just bogus and I never ever really got the point of them.
Seriously, if I have a company and need to decide what software I want to install on my computers, I’m not going to check and see what my prospective competitors use but what fits my demands best. As simple as that. I mean, if someone jumps out of the window, are you seriously going to jump after him or are you going to make up your mind yourself?
I really never understand how people in a free and educated world can be so much solely focussed on one big company and their marketing gibberish. Scrutinize your own needs and then decide. If it turns out that Windows will fit your needs best, then use it and thats absolutely okay. But not because Microsoft tells you so. Thats just the thinking and attitude people had in the medieval times before Kant came to the rescue.
Adrian
>>The truth is that all those statistics you report here can be forged by the click of a button. I’m using Google Chrome on Linux here. If I ever stumble across a website which is “optimized” for Internet Explorer, I just toggle the browser agent that Google Chrome reports to the web server and I’m officially a Windows user.<<
How often do you think anyone bothers to forge a user agent?
How many people do you think know how to do that?
Do you even know how to do it?
The lone user agent switcher extension for Chrome comes with this caveat: “sites that sniff user-agent strings at the network level will see through the spoofing.”
There is also the reality of OS specific downloads: The graphics driver for your video card.
Every time they stumble across such stupid websites. eBay is a good example. It hasn’t been working properly for a very long time because it did not properly detect Debian’s Iceweasel so you had to switch it to “MSIE 7.0” to work, even though Iceweasel was able to render the website properly.
That was not the point. You don’t seem to get it. They point is, it is very easy to forge it. When you start making statistics you should always be careful about the credibility of your results when your input data comes from very unreliable sources. This is something you always have to consider in science, always.
Ok, now you’re starting to make a fool out of yourself. Of course, I know how to switch the user agent ID when I told you in my last post that I have been using that ever since such “optimized” websites arose.
Ah, good point. That explains why forging the user agent doesn’t work in some cases with Chrome while it does with Firefox. Thanks!
This is, if you read the description of the extension more thoroughly, a limitation of Google Chrome. In general, the user agent can be easily forged. I can even use “curl” and make it identify itself as a browser running on Windows. “curl” doesn’t even have a GUI 
.
That argument goes well down the toilet. You’re ignoring two extremely important facts:
1) Linux drivers, even the proprietary ones, (as with most of the software you are using on Linux) are shipped out by the distributors and not the manufacturers of your graphics cards
2) Drivers for 90% of graphics cards are already included in the X.org server; there’s simply no need to download them separately
Let’s ask the question the other way around. If Linux’ market share is really that low, why do you think do companies like Intel, nVidia or ATI even care writing drivers for Linux? Those companies depend on profits, they certainly don’t do that because they think Linus Torvalds is a nice person.
There also a lot of prominent examples of large enterprises running mission critical software who have recently switched to Linux. Two of them which come to mind right away are Tokyo and London stock exchange. So do most universities and especially in the natural sciences like physics departments.
Adrian
So .. the number of folks using Iceweasel on Debian that were *not* counted as such on eBay over the past 10 years …
is probably 1/2 of 1%, compared of IE users. In other words, irrelevant.
You desktop linux nutjobs never learn.
Edited 2010-06-06 01:22 UTC
What kind of math is that supposed to be? “1/2 of 1%”?!? Is it too difficult for you to actually multiply two numbers?
And please explain me what “0.5% compared of IE users” (sic) is supposed to mean? 0.5% of IEs market share of 60%? That would be 3% total market share.
You obviously don’t know that in Europe Firefox has closed the gap to IE’s lead long time ago. And in Germany is has even surpassed IE.
http://gs.statcounter.com/#browser-DE-monthly-200905-201006
Adrian
What kind of maths is that supposed to be? Is it too difficult for you to actually multiply two numbers?
I think the number you are looking for is 0.3%
W3schools seems to be targeted towards people who are just starting to learn how to do things with html/js/css. I wouldn’t expect many hardcore linux users to go there. In anycase, I wouldn’t expect many linux servers to have a gui installed that would allow them to navigate to such a site.
We are talking about servers, right? I’m really not sure what the relevance of talking about desktops, when discussing server market share.
I would think that something like netcraft would be a much better gauge of server market share. Or at least Webserver OS market share. My company has never purchased a server with a preloaded OS: they all run Linux. Take a look at the really big websites. What do they run? Google: Linux; Facebook: Linux; Wikipedia: Linux; Digg: linux; Twitter: Linux;
The only large non MS owned site to have run Windows, was MySpace.
Where do all of those windows servers go? I think they’re mainly used for Exchange, Active directory, & Sharepoint servers. In other words, inside large companies. I think, this is just a very biased guess, the web server market is probably growing faster than the interoffice server market.
Am I biased against windows, yeah pretty much. But, there are some good historical reasons for that. I started out as an all MS guy, before I discovered the beauty of the Unix methodology.
W3schools seems to be targeted towards people who are just starting to learn how to do things with html/js/css. I wouldn’t expect many hardcore linux users to go there
I take exception to that. It’s an excellent resource for quickly looking up object APIs for JS, and all those pesky CSS attributes (and which one’s work with what browsers, etc.)
I keep it bookmarked for just that reason.
I can’t be bothered to remember everything about every language I’ve ever learned. That and I’m going senile.
Net Applications stats always come out different than a lot of others. My guess is Net Applications is measuring a lot of business users.
Then again Microsoft is one (yes of many !) of the partners of Net Applications. 😉
They use their own regional scaling system that they don’t fully explain. The problem with sliding scales is that you can make them say anything.
Edited 2010-06-05 19:33 UTC
Especially a sliding scale that you don’t explain and when you don’t explain your collection metrics. In other words, statistics without explained methods are completely useless.
Which has nothing to do with server usage.
He says: People can buy OS-less machines * and install free versions of Linux, while MS can count both the OEM copies and the boxed copies and you reply by saying Windows 7 has 25% of the Steam market share.
You do understand that servers are not used to browse the Internet – let alone play Steam games, right?
– Gilboa
* My previous employers bought 1000’s of OS-less servers and installed custom Linux, CentOS and RHEL on them. Needless to say these machine were never registered as Linux machines in surveys such as this.
not even the RHEL ones?
No, because in terms of sales they were registered as RHEL licenses, not as servers with RHEL installed on them.
The RHEL ones were under what-ever-support-license we had at the time.
Most of the OS surveys are based on shipped servers + OEM license. Support licenses are not included.
– Gilboa
Edited 2010-06-06 13:32 UTC
Steam users generally do not run it from servers.
Those are for desktop use, not servers so it does not really apply at all here.
Again, we’re talking about servers, not desktops. You seem to be unable to tell them apart.
Well, in that case, we would have to count all the pirated windows server too.
As the guy says, it’s already quite a feat to ‘sell’ Linux anyway.
But to put the numbers somewhat in perspective: I have some difficulty finding the original data, but according to the linked article in what was Microsoft’s best quarter 1,379,487 units have been sold. Let’s (for argument’s sake) say that on average a Windows installation runs 5 years before being replaced or upgraded. 5 years x 4 quarters x 1,379,487 units make a maximum of approximately 27,5 million running currently Windows servers. Since this also includes (a lot of) volume licenses, I take it the actual number must be lower.
According to that nice overview a while ago ( http://gizmodo.com/5517041/googles-insane-number-of-servers-visuali… ) Google owns with ~1 million servers about 2% of all the worlds commercially operated servers. That is about 50 million servers in operation. Many home and SMB owned servers won’t be included in this number, so the actual number is probably higher.
If I compare both numbers, I would suspect that at maximum 50-55% of all currently operated servers are running Windows.
These numbers also suggest that only about 2% of all Linux servers are running some commercial or commercially supported Linux. That seems to be a realistic number in my experience.
I don’t think home servers count as they are not remotely within the same market segment–plus in that case you would need to include Microsoft Windows Home Server in the stats as well.
I think these generalized figures from IDC are kind of pointless though anyway because they lump all “servers” together. For instance, Microsoft Exchange, which runs on Windows Server, has 65% percent of the groupware market (source:)
http://www.ferris.com/2008/01/31/email-products-market-shares-versi…
and I’d guess the figures are similar for Windows when it comes to file servers. Both of which make sense for the kind of Windows-centric desktop environment you find at the majority of businesses. Leave these two use cases out though and you will get a very different picture of things. When it comes to web servers, for instance, Linux and Windows are neck and neck at about 41% each (see the Netcraft figures in the following link):
http://en.wikipedia.org/wiki/Usage_share_of_operating_systems#Serve…
Also in that link note that Unix servers have around 30% of the market by revenue even though they have only 4.4% of the market by units. This again makes sense IMHO because I’m guessing that a single “unit” of one of these Unix servers performs the same amount of work as a bunch of x86 servers combined, or performs a very specialized, high-reliability function, thus justifying the high cost.
All of this just goes to show that the word “server” can mean any number of things, and statistics that lump them all together really aren’t all that useful.
Edited 2010-06-06 13:05 UTC
Quit been such a sore loser. This isn’t some 3 pence do it yourself crap we are talking here. Most of servers unlike desktops are available with large variation of OSes preinstalled and all have usually paid support available(often offered by Red Hat or Novell). Why you think these companies are doing rather well. Good example is in my work where majority of our customers used to have own internal IT to take care servers but nowdays almost all have outsourced servers to companies. And that means there is less “I heard Linux is great or I use Windows to play games so I’m expert” type people messing around systems. It also means they buy systems with OS on and not fiddle around like some god damn amateurs. Ofc what is missing on these numbers is virtualization, this could mean there is tons of servers sold with Linux but running bunch a Windows instances.
Not everyone wants “paid support”, and those who do usually have separate supports contracts rather than purchasing individual copies of RHEL or SuSE.
Why does outsourcing your server support or services to a third party preclude the possibility that those servers might run Linux? If anything it increases the possibility, because using Linux will reduce overheads which means more profit for the company providing the service.
Yeah ‘cos all Linux admins are just playing. Put me in front a Windows computer and I can’t even grasp the mouse. It’s a wonder I don’t have a seizure.
More likely there are a bunch of servers sold without an OS running VMWare running Linux. Like the 300+ Debian & CentOS virtual machines that I admin. That’s not even a big installation by VMWare standards.
Somehow the word “bullshit”, is an understatement in this case.
In my experience a sane IT manager will -never- use preinstalled OS for servers, it’ll use their own customized OS image.
I spent the last 15 years working in small to large IT companies (as a software developer) and not -one- of them used preinstalled servers, not for Linux and not even for Windows.
While the Windows servers were either covered by Microsoft site or bulk license, the Linux servers were either covered under some type of support license (E.g. RHEL) or supported in-house. (Mostly custom Linux built around CentOS and Slackware).
… Heck, in most of my previous workplaces, even the desktops and laptops (!!!) used a some type of heavily customized image of Windows or Linux.
Just in-case you think I’m simply making things up, I’m current working on a project that will be installed on (hopefully) 1,000’s servers running some type of customized CentOS 5.4 OS. Needless the say, the client is handling the installation and support in-house.
– Gilboa
Edited 2010-06-06 14:02 UTC
This is a nonsensical argument. Pretty much everybody who buys servers also buys a service/support contract with the OS vendor or hardware aggregator. It doesn’t matter whether it’s Red Hat, Microsoft, Sun, etc. So, these numbers corellate with actual market conditions.
Microsoft has got good marketing and supporting division when compared to Linux.
“Linux” is not a single product that can be compared to Microsoft’s products. If your going to compare marketing and support departments, you should focus on specific distributions. Given specific distributions, MS does have far better marketing but we’d need to have proper data on support from Novell, Red Hat or the various third party support providers.
Linux as a whole needs to band together to get even 1% share. If you break it down to distributions and then compare that against Windows, it looks even worse.
That’s not even a fair comparison I think.
If you want to compare distributions then you’d need to compare that against specific versions of Windows.
Putting all that aside I think all of us around here are well versed in the fact that comparing percentages on this topic is pointless and just causes arguements.
If an OS works for you then just use it I say.
Because most of the saner readers here know too well that such statistics are good for exactly one thing: nothing. They can’t be trusted, since the way they are gathered, where and how they are counted are most of the time unreliable and as always, every similar[ly unreliable] statistics can be presented in almost endless ways to favor one or the other component.
So these statistics aren’t showing a trend away from non-x86 platforms?
It’s sales data, it shows that sales of both Windows and Linux servers are up, it isn’t as if they are statistics being pushed by a political group.
Linux has always been better at displacing Unix than Windows Server, so these trends really shouldn’t be a surprise.
I disagree about the status of Microsoft´s server platform. I wish the editorializing would stop and the figures would be presented so that people can make up their own minds.
Just yesterday I had to move a Linux installation from an aging server to a new one. Different motherboards, different processors, different chipsets and all I had to do was move the hard drives over and watch it boot as if nothing had changed. My fstab mounts the drives by uuid which helps.
Today our windows server redundant power supplies took a dive and HP won´t send new ones until Monday or Tuesday. But, even though we have other hardware where I could move the drives of those windows servers, it would be useless as they simply refuse to boot. If they did boot, I would have to reactivate with Microsoft, Autodesk for our License Manger for Autocad Civil 2009, etc.
So both through technical shortcomings and social ones, proprietary software companies make my everyday life difficult as a sysadmin. Which is why, with every passing day, they get less and less of my company’s money, whereas we are happy to pay for many of the floss software we use, even when, many times, we don´t really need the support.
You do realise your comment proves people can make up their own minds, right?
*sigh*
While I’ve had similar experiences on numerous occasions and just been able to take just about any machine I had sitting around to temporarily fill the role and keep them running the reality is that for most small businesses this isn’t really an option. Windows servers pervade this market because they are relatively simple to set up and run, and provide lots of tools for the user to make it fairly simple to run their own stuff.
At a personal level I wouldn’t touch one with a barge pole because I know what I can do with a Linux box, and for those clients who are happy to pay for their systems to be maintained by someone else Linux is an excellent choice, but there are many businesses where this simply isn’t an option either from a cost or logistics perspective, so Windows will always win out in those situations. Of course there are also some applications that are tied to proprietary Microsoft technologies for which you must have a Windows server.
But again, for Linux to succeed Windows doesn’t have to fail. There’s plenty of room out there for numerous server platforms, just like there is for desktop and smartphone platforms. Ideally it would be like a herd of Unicorns, all playing happily with each other every day…
I don’t know how that started but Unicorns it is..
Anyhow, I’d like to see Windows play better with others so that my herd of Unicorns could happily play together. Right now, there is still to much bending the Lin Unicorns over backwards to mostly inter-operate with the Win Unicorns. Too many Win Unicorn choices impose further Win Unicorn choices still.
+1 ! Does someone remember how the Unicorn story began ?
Yeah, I want to know too. Don’t think I was hangin’ round here whenever that started, so I’m completely missing the joke.
OK, what is missing in Linux to make it easy to run and keep running ?
The whole notion that windows is relatively simple to setup and run is pretty much down to marketing… Sure, someone with little or no experience can get a windows box to perform typical tasks… But a modern linux distro or macosx server is no more difficult, users just haven’t been conditioned to think they could manage it.
However, the end result of such people configuring and running servers is horrendously poor security and stability… You could walk into virtually any company – especially very large ones, plug in to an ethernet socket and with a handful of freely available tools very quickly compromise the windows domain which is invariably present and take control of everything on the network.
Don’t larger companies have different teams for server administration and network administration? If so you cant really blame a junior serveradministrator for being able to plug in the network everywhere. You should’nt be able to gain access to the network without unplugging other stuff first. Best practice and security policies and such…
But to follow your logic they wouldn’t be able to plug into any ports if only the servers were linux or OSX.
As one of many who has been deling with both Windows and linux i’ll have to blame the administrator for unstability on a windows server. I have yet to see a windows server that arent stable as most linux systems i dealt with in the past.
Let me tell you that using Windows 7 RTM I had the exact same experience. I pulled the HDD out of an old PC then put into an entirely new upgraded PC and it worked immediately once it self installed the drivers. It took about 5-10 minutes to get them all installed and I didn’t have to provide any drivers.
I went from a AMD Athlon64 3200 on a MSI mainboard to Intel i5 850 on an Gigabyte mainboard.
On the other hand, there’ll be plenty of others who will cry foul at both of our experiences because they got a kernel dump or blue screen when they tried.
Both OS are very capable. This comparison doesn’t make either more superior than the other.
Agreed. I’ve had both Windows and Linux fail when drives were moved to different hardware. Fedora 12 failed to even load the initrd due to the sata controller having a slightly different configuration, and it had to be fixed from the live CD. I’ve also had XP and Vista fail when moved, typically they fail to boot for a similar reason (either different disk controller hardware or a different configuration of that hardware). Neither os is better than the other in this case. If you want to see an os that makes moving it to a different hardware setup a bit of a pain, try that with Solaris sometime. In either case though, if one doesn’t know how to recover and get the system back to a bootable state in the event of a hardware change, I think they’re in the wrong job position.
However, one must admit that the windows activation concern is valid, even once you do get it to boot. Unless you’ve got a volume license key, you’ll have to transfer your activation (assuming you have a non-OEM key and are allowed to do so). If you do have an OEM license, you’re sol unless you want to crack it illegally (many do). That’s one thing I seriously think MS needs to rethink from the ground up, their activation isn’t tolerant of much hardware change and can be a real problem if you end up in a situation where you need to transfer your installation to another machine in case of failure (that includes motherboard replacement).
I’m a Solaris admin in a enterprise environment so yeah I can agree with the Solaris comments.
We’ve never had a problem with Solaris in this regard. All the drivers are included, the system even lofi-mounts the correct libc version on it’s place. All you need is a reconfiguration boot. And if you want to get fancy, you can use flar or ufs dumps.
I hope you paid for that Windows 7 license on the second machine – otherwise you are breaking the licensing terms (if the first copy wasn’t already a pirated copy).
Not necessarily. It depends on what kind of license you have. Most Windows licenses allow you to transfer that one copy of Windows to any one machine, i.e. you can move it to a second machine and activate the new system which will transfer the activation to that system and deactivate the machine it was moved from. There are OEM licenses, however, that do not allow this activation transfer. I’ve never seen an OEM license like that for server though, just for desktop. In either case, you’ll know when you try to activate the new machine whether you are allowed to transfer or not.
Less than 10% percent of licenses are transferable. Depending on the license he has he probably won’t be able to transfer it. And certainly not for servers (which is the topic of this thread after all).
Um, no.
All he has to do is call Microsoft, and they’d transfer the license. Takes all of four or five minutes on the phone.
You lintards are incredible.
transfer? nope, Windows licenses are not generally transferable to new computers (sure, some are but the majority are not). Requires new license for ‘new’ computer. Upgrades perhaps, but depends on what MS says a ‘new’ machine is.
Plus, your name calling is very petty. Please leave it for the junior high playground. I never once breathed a word of Linux in this discussion so take your discrimination elsewhere.
Edited 2010-06-06 03:55 UTC
oh good grief, calling you a “lintard” is now “discrimination?”
I don’t suppose there’s a chance that you’re a bleeding-heart liberal egghead, is there?
Get a backbone. You personify the pussification of modern secular humanistic society.
And yes, if you call Microsoft THEY WILL transfer a license to a brand new mobo/cpu. I’ve done it.
Edited 2010-06-09 15:05 UTC
FWIW, I needed to do this on a W2K3 server just yesterday. It worked fine.
I moved the array, plopped it into a new, entirely different server and it booted fine. It did want to reactivate, but that takes about 2s.
I’ve got one 2K3 server install on its third box, each one with totally different hardware.
I’ve had the exact opposite occur, where I’ve moved a Windows server to a different box, and it re detected everything just fine, and I know if I moved some of my linux servers to new hardware, they would fail, it’s not proprietary versus open, it’s just the vagaries of configuration.
Every server is different, and a lot of times, it’s just one weird driver that keeps it from moving smoothly.
People shouldn’t look at things where Windows needs to fail. As long as Linux is growing, that’s a good thing.
The main thing is, we need competition between products and that’s one reason why Microsoft’s products have become better as Linux tries to become better itself as a product.
One day I hope to see the same for the desktop market, where both can compete on an equal footing.
Windows is behind in the HPC arena, and that’s all I care about.
Window’s HPC server isn’t bad, I must say.
I’m glad you included the ‘pre-installed’ software reference in your article. But is this article really necessary? Server sales info like this is primarily PR stuff, and not an accurate gauge of actual, real-world presence. Stats on what operating systems are moving data around online and in various companies are really matter.
As if we didn’t already know that this whole thing is a click through scheme for computerworld…
Hopefully I just missed it, but has anyone considered how virtualization affects the marketshare numbers? Right now it’s the big thing in data centers and cloud efforts. It’s entirely likely for any of these systems to be running multiple copies of Linux or Windows on top of it.
Edited 2010-06-05 00:59 UTC
It’s very easy to make Microsoft look good in these. Every single (ok, almost every single) license is documented.
For Linux it is completely different. I’d say at least a good 70% of Linux servers are not documented in such a way that you could count them in sales.
For instance. Where I work there are more Linux servers than Windows servers. However, the computers were bought without any OS on them and Linux was installed later because OEMs don’t sell the configurations that our server team needs and they have images that they load onto new servers. Why pay HP or Dell or any other company for configurations of software we won’t use.
So … NONE of the Linux servers, even though they out number the Windows servers, get counted. There is no license to buy. You just take the same image you have and copy it onto server after server after server.
Where they SHOULD get their stats is based off of OS updates for when servers connect and get updates to their prospective OSs. Except for one thing. With Linux it is far easier to get one copy of the updates, verify that everything is working, and then patch all the other servers with those updates and configs. Again, very little show for how many copies of Linux are being run.
Support? There is far less issues with Linux servers Once you figure out what you are doing with them. Once you do, a lot of the pain of servers (compared to Windows) goes away.
So again, these kind of stats are BS. And the BS are in CAPS.
Home repositories is a good one for skewing statistics when they are able to track distro repository downloads. I used to have several Mandriva systems running here and all drawing from the same repository mirror rsync’d down to my local NAS. Why pay my ISP more than once when updating several machines or rebuilding some testing system when I can download once and host the repository inside the LAN?
(and, I really have to add that it’s fantastic to be able to host a complete repository locally without licensing issues to worry about.)
I tend to agree. Especially since the web server market is relatively tightly controlled, so just checking DNS logs would probably give a much better picture of the thing.
That is for web servers. Such stats would have no value about company servers which run a private intranet and do not need to be connected to the internet. The parent poster provided some insight about how to check things in that areas. Those two sales figures should be separated, in my opinion. These are different markets, and comparing them would be interesting…
Edited 2010-06-05 07:05 UTC
Red Hat and Novell are not building their distros with the typical business in mind. They just sell vanilla Linux distros with corporate support.
Setup file and print sharing in both Cent and Windows Server and you’ll see why small and medium sized businesses have no problem writing a check to MS.
Ok, I can do that on Cent in under five minutes, but I haven’t used windows Server in four years now. I think it would take me longer than five minutes.
What was your test supposed to prove?
Maybe for new non technical users? I still think that’s about equal time, if you put a gui on Cent.
Now, throw in Exchange like email capabilities, or an existing SharePoint App, or Domain controller and Windows wins hands down. That is the real differentiator for Small and Medium Businesses.
You’re not the targeted demographic. The typical business owner does not have experience in Linux and there is no push by a company to make it easier for them to adopt Linux servers.
If they learn how to setup a GUI, if they don’t have problems with CUPS, if they don’t have problems with Samba, if they don’t have problems with printer drivers (multi-function printers are still a problem in Linux), that’s too many IFs to be worth the initial savings over SBS.
Novell or Red Hat could address these issues but they focus on the corporate market.
Actually setting up Windows SBS is beyond most, if not all, people that tend to buy into it.
I’ve been proven time and time again, that those SMB owners see “Microsoft” and “Windows” and think they can do it easily. Since, “You know, I have Windows on my computer and it can’t be harder than using that….” But the fact is, that they can’t*.
It’s as hard as setting up a Linux server for them, because they still need an admin to come in and do the work.
*- You can’t imagine how many times I’ve been called out to help those poor gullible bastards, that didn’t consult an IT person prior to their purchase.
That you’re an idiot?
How many small and medium-sized business owners (who KNOW how to use XP, Vista, or 7) would have any desire or inclination to try to do the same thing on CentOS, Fedora, Ubuntu, or anything else?
Linux is fine for higher-end servers. Pointless for desktop and small business usage for *anyone used to windows* … i.e., the millions and millions of computer users that would never post messages on this site.
Infuriating simpletons, the lot of you — completely unable to grasp reality.
Edited 2010-06-06 01:36 UTC
Woah, put the blowtorch and cheese away my friend. This isn’t a fondue, nor a flame bait contest.
Did anyone ever counted the vast amount of servers running Debian? I do not think so. This is just MS PR, none of the servers I installed was ever counted.
Users don’t care what you post, whether it’s Microsoft, Apple or Linux.
It’s the ways you support certain things, when the next company does the exact same thing and gets slammed for it. Even more frustrating when there’s no real world experience in the area you’re slamming.
That being said, I’m sure Windows share is improving again, they’ve done some solid releases with their servers and I won’t take that away from them. Just like Vista was a disaster but I hand it to Windows 7.
It all depends on the area you’re trying to service. The majority of the times I find a Linux or Apple server far superior to the Windows servers in stability, ease of upgrades, maintenance.
I think the reliability issue has become somewhat of a canard. Linux, FreeBSD and Windows Server are all plenty reliable. Linux companies can no longer rely on stability to be a selling point. It kind of reminds me of Sun who was dismissive of Linux and just repeated “reliability of Solaris” as a selling point even when it was clear that Linux based web services were plenty reliable.
Windows 7 share is certainly improving – at the expense of the Vista and XP for sure.
I think it is unfortunate that people use numbers like they tell a story.
True, microsoft is making money regardless. If I were going to get a server, I would buy the server I want, with windows on it, because the linux they offer on the server isn’t the one I use, and basically to get it only cost me the price of a CD or DVD. I think a lot of people who have this choice to make would make the same decision.
Where are you buying servers that come with a free copy of Windows? I’ve never even seen the $400 Web Edition discounted.
Junkware on a desktop can offset the price of Windows but on servers you pay the full price.
These server sales surveys really don’t tell us much except whether Microsoft is getting more or less revenue for its OS being pre-installed on servers. You *cannot* compare Windows (which is 100% pay, except for those who pirate it) and Linux (which is primarily free, except for larger companies who might want paid support).
Remember that most server OEMs now offer an OS-less server (I wish they’d do that with netbooks/laptops/desktops too!) – we buy mainly Dell and HP servers at work and both are easily shipped with no OS at all (and, except for companies that pirate Windows, those will typically have Linux installed on the vast majority because buying Windows Server “retail” is usually much nore expensive than buying it OEM pre-installed).
So with most (all?) “OS-less” servers having Linux (probably CentOS or Debian, since they’re the most popular free server Linux distros), shouldn’t such servers be counted as server sales for Linux (well, most of them anyway)?
Sadly, we’ll never really know the true market share of Windows vs. Linux on servers, since it would require Microsoft and all major Linux distros to produce figures on the number of unique Windows Update/update repository requests made by the OS (even that isn’t accurate – e.g. local update servers or proxies would muddy the figures), but it would be a lot more accurate than this server sales guff we seem to get year in and year out.
Well… when the article is about server OS marketshare then why does everyone discuss about laptop/desktop OS marketshare? In that sense status quo has been the same for years – 94% Windows, 5% Mac, 1% Linux (all of my PC’s are in the last segment).
However on server side I believe the same what is mentioned in the article – a lot of Linux selfinstalled installations are uncounted and it is hard to find total figures (e.g. Netcraft shows only the webserver market structure).
The small business I work for runs 3 bought Windows servers and 25 selfinstalled Debian GNU/linux servers (i.e. bought as Dell server hardware with no OS). And all new installations during last 5 years are Debians
Edit: fix typo
Edited 2010-06-05 19:32 UTC
Plus, what about the *growth* area of the market? – ‘smart’ phones. That’s where there is huge money to be made and Windows share there is plummeting.
Microsoft is also stagnant yet again in the share market – their balance sheet is ok but not as strong as it used to be several years back.
The whole article could well be a pep-talk for investors who are otherwise shaken.
It very much depends on exactly what you count, and how you count it, but overall Linux share is way, way beyond 1%.
http://www.desktoplinux.com/news/NS9682048723.html?kc=rss
Net Applications reports are known to significantly under-count Linux. There is a large base of self-installed Linux out there (machines like mine assembled from parts, or ex-Windows machines rescued from the tip after a death-by-malware-and-registry-clog only to be re-birthed running Linux
http://linuxagainstpoverty.org/
http://www.untangle.com/Products/Installfest
http://installfest.org/ ).
It also depends on exactly where you gather your stats from, or which exact market you survey.
http://www.desktoplinux.com/news/NS5247302500.html
Still, even then, counting only new-machines-purchased-from-controlled-shop-channels-in-western-countr ies, Net Applications gives Linux credit for 20.8% servers, 1.13% desktop, about 30% mobiles/phones, and 91% of supercomputers.
This is a significant slice of the OS market share, and far, far more significant that Windows supporters would paint it.
Edited 2010-06-09 12:38 UTC
If Microsoft wants to throw number around, lets count the number of linux devices in the world. Lets include computers, tivos, routers, and all the other devices that run a linux kernel. Microsoft could only dream of have a market size that big!
Centralised authentication, useful filesystem ACLs, a network filesystem that doesnt require total lockdown for any client accessing it,GUI tools for management of the above – all of this is missing from Linux, and present in Windows.
You can do some of those things on Linux, but only in a way that requires huge amounts of arcane knowlege to
set up and maintain. Some fundamental stuff, like filesystem permissions, will require enormous, pervasive changes to many layers of tools to work nicely.
However, When you can do the following in Linux using the native, gui, non-samba management tools supplied out-of-the-box by a distro:
Add 2 users to the system, and create 2 groups.
Add the first user to the first group, and give the group read and write access to a folder.
Add the second user to the second group, and give that group read-only access to the same folder.
Share that folder across the network
Give User B a home directory on the server.
Share a printer across the network as Printer A, and give print access to both groups created earlier.
Have User A change their password when they next login.
Allow User B to log in from their personal laptop, which uses a different set of uid/gids to the company server.
Set a print quota for Group A so they can be billed per-page from Printer A
Back up and restore the files on the system such that all file permissions are retained.
Then a lot of people who are dependent on Windows will sit up and take notice. Until then, that’s why Windows server is popular.
Edited 2010-06-06 00:28 UTC
All these things can be done NOW, with SLES.
Wow, it looks like that might actually be the case.
Its almost like Novell is dragging Linux, kicking and screaming, into the 1990s.
I’m sure its actually not quite as simple in the real world (especially where users changing POSIX permissions meet NFS4 ACLs), but SLES11/SLED11 looks like a pretty decent NFS4 server/client.
This is all sadly true.
What Linux is missing is not capability (not technically) but this kind of integration and usefulness. Why oh why is it so difficult to set up a Linux network with integrated authentication and file exchange? Why does every single Linux box still insist on defaulting to such “what network?” configuration?
I know UNIX was designed pre-network (and it shows!) but surely there is a way, today, to make it all work nicely and *easily* in a way that requires no hand-tuning or HOWTO-reading.
Is there any distribution that lets me set up a windows-like domain (meaning kerberos+ldap) out of the box? How about one that lets me “join” such a domain with nothing more complicated than a few clicks? I don’t need this myself, but it’s a damn hard sell to most people that you need someone like me to make it work.
There are distros like ClearOS that try to replicate Windows SBS, and is a quite useful OS for really small businesses in static markets. However, as I tried to point out earlier the differences between these two operating systems is night and day (at least for me).
I know that you can set up Linux to use Samba+LDAP to create a domain environment, but what environment have you exactly set up? Something akin to 1990s NT4. While that may suffice for some smaller shops, it simply does not cut it for the vast majority. Active Directory domain services in 2008 is far beyond what you can do anywhere else. Couple that with Group Policy Management and AD Rights Management and there simply is no alternatives. Through in Terminal Services and you really have no choices.
The same can be said for Windows in regards to those that wish to set up large storage networks, multi-node clusters for dB, web, virtual machines, etc.. Do many run Windows Server in large data centers of VMs? Not many (not a single one that I am aware of).
Just because you can get an OS to do something doesn’t mean squat. I can get my car to drive backwards at good speeds, but I sure in the hell don’t drive down the road backwards………often that is.
This is simply a metric of what is being done in terms of sales, not usage. Yet there is a complete lack of reality amongst most. This entire thread is nothing but mindless dribble of excuses or denial.
Fact of the matter is Linux and Windows Server rarely compete in the same space. A more precise metric would have been figures based on data center sales.
What these figures do represent are servers, no specification. Therefore one could easily apply the logic that Windows server sales are up simply because businesses are spending less on data center growth, while many companies have been upgrading hardware along with Server 2008/R2 migration.
Again, these are operating systems that are employed for mostly different reasons. Multi-node clusters, data storage, dB clusters, web servers, VMs, etc.. fit a much different role than Active Directory, AD Rights Management, NAP, Terminal Services, and most importantly Group Policy.
This like whining about your favorite football team losing popularity to Oprah.
I guess they mean of all servers with bundles OS license, Microsoft Windows share grew with 4.7%. Personally, I buy my servers without any license since I’ve a hardware vendor independent contract with both Microsoft and RedHat…besides, they’re VMware Boxes anyway.
I’d like to know how many units were sold in total which gives a percentage of units sold with bundled OS. I.e. leave the paid conclusions be and dish out the raw data!
What I am more interested is to know how many actual servers have been sold and how many have been sold without an OS. Then we can make some guesses how much new servers will run Windows.
Because this statistic, though probably true, shows only the market share of sales of complete systems. Witch will definitely go to Windows, without surprise.
– from the article.
Thom, you could post the current time of the day as a story and half comments would complain that your pro-Linux, pro-Microsoft and or anti-Apple somehow.
I always find it funny when a bunch of Linux admins try to tell Windows admins how hard Windows is to use. Or when Windows admins tell Linux admins how hard Linux is to use.
I feel like finding a Chinese person and explaining to them why Chinese is so much harder for ME to use than English. Lets get together and we can all pass out FAIL stickers to each other?