Re: It's what you might call a limited offer....
Actually, I deal with "the mainstream market", being "the 80% of businesses" represented by SMB and SME. That, however, is neither here nor there. I made no claims in this thread whatsoever as to who has the most servers.
What I do notice, however, is that a historically rabidly pro-Microsoft commenter has cherry picked one set of statistics to claim dominance of Microsoft across "servers", then backs it up by saying "even if the stats aren't right, my personal experience verifies them."
What that tells me is that you are allowing your personal experience and personal preference to prejudice you. To the extent that you will cherry pick for stats you like and then hold them up as gospel. That makes absolutely anything you say unreliable, whether it coincides with my personal experience or not.
Personally, I have no idea what the ratio of Linux to Windows is, unless I have an exact definition for server. And you can change the definition of "server" in a lot of ways to get a lot of different results. I notice that you picked one that favoured your personal preference that Microsoft be viewed as dominant.
Me, I'd look at embedded devices serving a "server" function as well. Each node in a storage cluster would be counted. I'd count hypervisor hosts as well as guests. I'd even count routers. Anything that is a computing device but isn't actively used by an end user for input and response.
In my personal experience, that would put Linux/Unix/BSD as the foundation for around 120x as many systems as Windows. That drops significantly to about 35x if you count client OS VDI instances as "servers" (and frankly, I do.)
Now, how does my "personal experience" rank against the world population of servers by my definition? I am as yet unsure, in large part because a lot of embedded stuff simply isn't listed. Another part being that a lot of embedded stuff is VXWorks or QNX, which I wouldn't count as Linux/Unix/BSD, but make up a huge percentage of deployed systems.
Moreover, I come from an Oil and Gas province with a startling number of high-end post-secondary institutions and which is a leader in nanotech; there are lots of reasons why Linux/Unix/BSD is more popular here than it might be elsewhere.
So I simply don't have enough info to make grandiose claims about how many "servers" are Windows and how many are Linux, ever if we could ever agree on a definition. The stats I have seen, however - and I mean both locally and globally - do seem to indicate to me that by my definition there is a fuck of a lot more Linux/Unix/BSD out there than there is Windows.
As a general rule, my experience says "systems where humans have to interact with the thing on a regular basis" are Windows and "systems that need to run for years at a time without intervention" are Linux. Windows is easier, thus it gets used for line-of-business servers that see frequent upgrades, changes, etc. Linux is harder (but way less fragile) thus gets used for things that absolutely need to work, or where the cost of rolling someone out to fix it/replace it is far higher than just doing it right in the first place.
But again, that's my experience. And the experience of the overwhelming majority of systems administrators, CIOs and developers I've talked to. I don't believe that is the final word because, to my knowledge, nobody is actually counting every system deployed. How could they?
What I do know is this: people I deeply respects choose both Windows and Linux/Unix/BSD for different use cases and they deploy both of these OSes where they feel the OS is best suited. In most places, my assessment of the when/when/how/why aligns with theirs.
"Total server deployment" statistics can tell me a lot about where global money is going, but they don't tell me which OS is "better" or inform me that I should use a given OS for a given use case. In fact, "total server deployment" statistics are really only useful for penis comparison and validation of one's own preconceptions.
What does matter to me are statistics regarding specific use cases. Here I fully expect that my local experiences will vary from the rest of the world: Alberta is not a province of inventors and innovators. We're a province that applies the knowledge and ideas of others to hauling minerals out of the ground or growing cows. We extend the work of others; we very rarely do anything radical.
Thus if there is a big misalignment of personal experience with statistics it could indicate that others have found a good reason to change which OS they use for a given use case, and it behooves me to check it out. To be ahead of the curve locally, not behind it.
So I return to the beginning of this post and say:
What I do notice, however, is that a historically rabidly pro-Microsoft commenter has cherry picked one set of statistics to claim dominance of Microsoft across "servers", then backs it up by saying "even if the stats aren't right, my personal experience verifies them."
You completed my commenttard Bingo for the day. I dobbed all the tropes on my card. More importantly, you have confirmed an opinion I've had for some time, which is that your personal prejudice is something you gleefully allow to influence your decisions, even when it comes to something as simple as selection of relevant and useful data sources upon which to base decisions!
My primary experience is certainly niche - though that is growing in diversity rather rapidly - but I don't generally say "my experience should be the basis of all decision making for all use cases." If I have a grouse with a company then I state what the grouse is and why I have it; if the grouse is "just my experience" I usually say "run your own tests to make sure."
If the grouse is something more endemic, like "how a company treats is customers/partners/etc" then I will give direct examples of what I believe bad behavior to be. Some times I'll relate an experience; this is usually followed by an invitation for others to share like experiences, that we me broaden and deepen the data available beyond just my own woes.
With the exception of an expectation that companies treat customers fairly and as reasonably, it has been a heck of a long time since I've said "one X to rule them all." IT is huge. It's bigger than the number of people on this planet! IT has become so diverse that it represents not only every facet of how every person and company chooses to do something, it now adapts to the needs of machine learning and evolution!
Anyone, anywhere who thinks that a statistics like "OS deployed on the majority of servers" means a goddamend thing is a fucking idiot. Anyone who claims they even have a means of tracking that today is a close second. You cannot even say "this is the best OS for storage/networking/cloud/virtualization/embedded/etc" because the major categories are so breathtakingly huge that diversity of requirements and use cases renders any attempts at broad generalization moot.
Your experience means fuck all, and so to you. My experience means fuck all, and so do I. The universe is so vast, and we are so small. The best we can do is deal with the individual use cases in front of us one at a time, and make the best decisions possible with the data we have to hand.
...and to guard against our own prejudices and preconceptions, lest they blind us to better options when attempting to navigate the most rapidly changing field of human endeavor ever undertaken.