Is Windows still the king of the server OS market? Absolutely.
The problem is that Microsoft's kingdom is a heck of a lot smaller than it used to be.
You might have a hard time figuring that out based on some recent coverage of IDC's latest worldwide server market numbers. Over on ZDnet, blogger Mary-Jo Foley sums up the conventional wisdom under the headline, "IDC: Windows Server still rules the server roost." And at Computerworld, Preston Gralla declared this week, "Windows widens lead over Linux in the server market."
Both blogs refer to IDC research showing that Windows Server widened its lead over Linux during the first quarter of 2010. The Linux share of the server market dropped slightly from 20.8 percent to 21.2 percent, while Windows Server's share climbed from 73.9 percent to 75.3 percent.
Peek behind those numbers, however, and you'll find they're a Potemkin village built on quicksand.
IDC bases its OS market-share research on factory shipments. If a server comes pre-installed with a particular operating system, it gets counted. If a server ships without an OS, it doesn't.
Guess which OS benefits from this methodology? If you answered "Windows," give yourself the rest of the day off.
That would be true even if most enterprises had never heard of server virtualization. Quite a few Linux and Unix-like server distros, including Debian, CentOS, Ubuntu, and *BSD, are downloaded for free and installed on bare-metal systems. And in many cases, commercial Linux distros like Red Hat Enterprise Linux and SUSE Linux Enterprise Server are also purchased separately from the hardware used to run them.
Server virtualization muddies the waters even more. Enterprises that free up server capacity through consolidation will, by definition, have to install new OS instances after the fact. Throw in the market impact of new servers running bare-metal hypervisors, and IDC's numbers get even less useful.
How does all of this affect the real server market-share numbers? I have no idea, and anyone who does is a either a psychic or a liar -- or is using different methodology. All I can say for sure is that it's ridiculous to pretend that IDC's statistics provide an accurate or useful view of the market landscape. Today, the company's research is nothing more than a quaint anachronism that quit telling a useful story years ago.
I don't blame IDC for clinging to its methodology. It would be economic suicide for any research firm to admit that its market-share estimates are a bunch of baloney.