So the latest IDC quarterly server report is saying that shipments of servers with Linux on them are on a downturn, compared to shipments of servers with Windows Server on them.
Game over, Linux, you are defeated.
Not.
(I have updated this article to make it more clear the point I’m trying to make.)
The main reason, that the eweek.com article, “Linux Losing Market Share to Windows Server” points out, is that the IDC survey numbers for Linux shipments on new servers are in decline mainly because the migration from older Unix systems to Linux-based replacements is slowing down, therefor translating into a downturn in demand for Linux on new servers.
Lets be clear about what the IDC survey covers. From their own site:
IDC’s 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
So basically, new server hardware and the OS that’s bundled with it. The survey is that narrow, and as such, it will be skewed out of all proportion to any real and meaningful usage figures. What I don’t seem to see is anything which mentions bare hardware with no OS bundled with it, because that might paint a completely different story to the one which is currently being painted by the likes of eweek, based on the IDC survey.
As an ex-IT manager, I know that I can buy hardware either with an OS or without. In my experience, we usually bought servers without an OS installed. Then we put Linux on it – Debian Linux to be precise, and Debian “Stable” at that, because we wanted to use a distribution of Linux we knew would be rock solid – the only package upgrades being security updates or bug fixes.
And it’s precisely that kind of Linux installation which will never register in IDC’s survey. And I can say this with absolute certainty that it’s this kind of activity which skews surveys like the one which IDC publishes completely out of kilter with the reality of just how many company servers are actually running Linux.
IDC even admit this:
IDC analyst Al Gillen pointed out that the number of servers shipped does not perfectly equal the number of operating systems in the market. This is particularly the case with Linux where a substantial portion of the overall market opportunity comes from deployments aboard recycled servers, PCs and workstations deployed as servers, and Linux deployed as a guest operating system.
But we do need to remember that the Linux software ecosystem does not track exactly the same as does x86 hardware shipments.
Indeed, it does not, for the very good reason that installing a Linux distribution on older hardware really does revitalise the usefulness of those old systems, saving companies shed-loads of cash in the process.
It’s also noteworthy to point out the increasing trend in installing Linux systems inside virtual machines, so you typically have a company buying say two or more nice big beefy servers, for example, then installing any number of Linux vm’s onto them. Someone pointed out that Windows installations are also being installed on these virtual machines, which of course is a good point and I add that point in this article for complete fairness. Are these Linux virtual machines being counted in any survey? Not that I know of. Please feel free to point them out to me if they exist.
If only there was a survey which counted the following;
New server hardware shipped with Linux
New server hardware shipped with Windows
New server hardware shipped with no OS
It would be interesting to see the figures on that, because I’m fairly certain, based on my own experience working in the IT industry, that new server hardware shipped without an OS has more than a 50% likelihood of eventually ending up running a Linux-based OS on it.
This brings me nicely onto a suggestion for the Linux Distribution makers; it would be nice if you could include a survey – doesn’t have to be particularly comprehensive – during the installation phase of your respective distributions, which could look something like this:
“Thanks for installing <name of Linux distribution here>!!!
We’d be grateful if you’d spend 30 seconds answering just a few questions on your use of this new system:
1) Purpose of this installation : (server, workstation)
2) Where is it being used : (company, home, other)
3) What kind of hardware : (raw metal, virtual machine)
And then ask if this data can be sent “home” to the Linux distribution makers. Just those three questions alone would provide, in my opinion, some very useful data on the number of Linux installations being used and what they’re being used for and where they’re being used. No need for identification of the type of company or person installing it, privacy of the user is kept (apart of course from their IP address), and if included as part of the installation process, it might guarantee a large percentage of the people installing the distribution make the effort to send the data back.
This data could then be published on the distribution makers’ websites, then we all could get a better picture of the situation.
I also wonder if the distribution producers could garner data on Linux usage based on the number of distinct ip addresses which access the respective distribution’s software and update repositories, though that would entail asking the numerous mirror repositories around the world to also supply that data, which is probably a lot to ask and is a lot of effort to collect.
I personally believe that more effort should be made to collect this usage data. It would benefit the distribution makers in that they would have real figures on just how many installations of their distros are actually out there, and it would benefit us too, for we would have a better and more factual basis for issuing a suitable riposte to the “losing market share” type of headline.
