Tea leaves and Linux market share

I was – just as the story lead suggested – a bit slackjawed over a statistic from Gartner Dataquest cited in a recent news story. The statistic causing that reaction was that the Linux server market share of U.S. sales had grown 90% in the 4th quarter of 2002 over the same quarter in 2001. Try as I might, I could not make that statistic fit with data from IDC.

IDC data, for example, weighs in with Linux server revenue in the first quarter of 2003 at $583 million, an increase of 35 percent over the same quarter in 2002. I am not saying IDC is right and Gartner wrong, I am only pointing out that they are different.

Tracking Linux market share is a black art, especially given the way that legally multiple Linux installations can spring from a single purchased copy, or from no purchased copy at all.

IDC and Gartner Dataquest are two of the most often cited sources of market share statistics. But they don’t produce the data in exactly the same ways. It’s sort of like buying a major appliance from Crazy Ed that comes with a guarantee that you won’t find it available for a better price anywhere else. Crazy Ed’s confidence in that guarantee comes not from their own low price, but from the fact that the appliance manufacturer has provided them unique model numbers that simply won’t be found for sale anywhere else.

Prior to 2002, IDC tracked market share on factory shipments of server platforms based on the number of servers shipped. That does seem a logical approach. But starting in Q1 of 2002, they began to track market share based on revenue rather than units.

Some in the Linux community feel that was done to make Windows look better since counting by cost instead of server gives Microsoft a three-to-one advantage over the upstart Linux. But the Unix segment is actually the big winner when counting market share by dollars: the Unix share jumped from just under 12% in 2001 (counted by units) to over 50% in 2002 (counted by dollars). This leap comes while the actual sales revenue from Unix fell almost 13%. As economists might say, go figure.

Another favorite measuring stick of market share is the long-running Netcraft survey of web servers on the Internet. Checking the latest stats from Netcraft shows Microsft losing approximately 10 percent of the web server market over the past year while Apache gained about 10 percent. Don’t forget that this metric is only concerned with the web server, not the underlying OS.

The Netcraft survey has not been without its own share of controversy over the way servers are counted: by domain name rather than physical machine.

The bottom line is that it is difficult to be precise about market share. It’s guesswork even before the additional complexities of counting open source usage is factored into the mix. But whether you believe the Gartner Dataquest claim of 90% growth or not, you can be sure that the Linux share of server market share is both significant and growing. That’s one thing everyone agrees on.

Joe Barr has been writing about technology for 10 years, and about Linux for five. His work has appeared in IBM Personal Systems Journal, LinuxGazette, LinuxWorld, Newsforge, phrack, SecurityFocus, and VARLinux.org. He is the founder of The Dweebspeak Primer, the official newsletter of the Linux Liberation Army.