My good buddy Preston Gralla would have it that "Windows doesn't just dominate the desktop, but the server market as well." Eh... I don't think so.
But Gralla and other critics are missing that IDC is not measuring what server operating systems are being used; it's measuring what server operating systems people are buying, and those are bundled with their hardware purchases. Specifically, to quote IDC, the researchers are measuring "server revenue includes components that are typically sold today as a server bundle, including frame or cabinet and all cables, processors, memory, communication boards, and OS."
Is it time to start throwing a fit and start selling Red Hat stock for a dime on the dollar, and enroll in MCSE (Microsoft Certified Systems Engineer) night classes? I don't think so.
Head to Head
What IDC is measuring is fine for looking at the traditional server sales market of Windows versus Unix. There, people tend to buy Windows or Solaris servers as a package deal. But that's not how the real data center and server world works today.
Today, many businesses buy generic x86 servers and then add Linux to them. Many of the companies I know have made Linux servers out of older servers or workstations. Still others buy cheap new x86 servers without any operating system and then put a free Linux distribution on them.
Take my own office, where I'm currently running five Linux servers. None of them came with pre-installed Linux. I installed RHEL (Red Hat Enterprise Linux), CentOS, Ubuntu, and openSUSE. None of these self-installed, small business servers will ever show up in IDC's study.
It's not just me and other Linux fans who are running their own pick of Linux distributions on ordinary x86 servers. Weta Digital, the company behind Avatar's spectacular special effects, used 35,000 cores on 4,000+ HP blades running Ubuntu to render the movie.
Guess what? HP doesn't sell its servers with Ubuntu. Heck, while HP has certified Ubuntu for use on its mid-range servers, it hasn't even certified Ubuntu on the HP BL2x220c blades that Weta uses to create special effects. In other words, here we have a gigantic example of Linux servers in use that IDC will never measure.
Besides, IDC's survey is all about measuring new hardware server sales. In this study, IDC isn't measuring server virtualization. So, for example, while IDC measures IBM mainframe sales, it doesn't measure how companies are consolidating thousands of hardware servers on System z mainframes running thousands of Linux virtual machines. As IDC system software VP Al Gillen told me a while back, "We do need to remember that the Linux software ecosystem does not track exactly the same as does x86 hardware shipments." He's got that right.
Oh, and one last thing: Even by IDC's measurements, "Linux server demand also improved in 4Q09 with revenue growing 6.1% to $1.9 billion when compared with the fourth quarter of 2008. Linux servers now represent 14.7% of all server revenue, up 1.4 points over 4Q08."