Robert Pogson

One man, closing all the windows.

The Rise of GNU/Linux on the Desktop

  • Feb 02 / 2012
  • 15
technology

The Rise of GNU/Linux on the Desktop

According to NetApplications, GNU/Linux continues its rise on the desktop. The trendline shows 8% per month increase in share, 150% per annum.

I don’t know what has changed in NetApplications’ world but in the real world, a rate of growth like that would make GNU/Linux the dominant desktop OS in 3.5 years. Android/Linux is on a more modest pace and will take over the world in 4 years.

15 Comments

  1. aikiwolfie

    I couldn’t give a monkeys about IDC’s numbers or anybody else’s numbers. For the month starting 11 Jan 2012 and ending 9 Feb 2012 my own little blog’s traffic shows browsers by pageview as IE with 32%, Firefox with 27% and Chrome with 26%.

    On the OS front we see Windows with 66%, Linux with 17%, Macintosh with 4%, Ubuntu with 3% and Android with 3%. Now Ubuntu and Android are Linux distros. Don’t know why Google splits them out. But it is interesting to note Ubuntu is doing just as well as Macs. The total for Linux is 17+3+3=23%.

    Which is actually a heck of a lot more than I should be seeing if estimates of Linux market share on the desktop/laptop are right and are in fact down around the 1%+ bracket. And it’s the same story for every website that records stats. Never have I come across a website that has Linux down anywhere near 1%.

  2. Mats Hagglund

    And at the same time W3Schools is showing Linux marketshare declining (4.9 % Jan 2012, while 5.0% in Jan 2011). Surely – i have never trust neither Net Application nor W3Schools statistics at all. That why i recommend everybody compare the statistics of Wikimedia (of course it’s too much “english language”, Apple have terrible too many “clicks”).

    http://stats.wikimedia.org/archive/squid_reports/2011-12/SquidReportOperatingSystems.htm

    Windows 3,360,079 74.20%
    Mac 388,666 8.58%
    iPhone 247,513 5.47%
    Linux 201,029 4.44%
    iPad 90,501 2.00%
    BlackBerry 32,374 0.71%
    SymbianOS 8,131 0.18%
    DoCoMo 711 0.02%
    FreeBSD 495 0.01%
    SunOS 390 0.01%
    OpenBSD 32 0.00%
    Total 4,528,429 100%

    Here are the figures of April 2009:

    Windows 3,367,634 89.50%
    Mac 227,797 6.05%
    Linux 55,999 1.49%
    iPhone 35,142 0.93%
    BlackBerry 5,598 0.15%
    SymbianOS 3,251 0.09%
    DoCoMo 764 0.02%
    SunOS 475 0.01%
    FreeBSD 372 0.01%
    OpenBSD 35 0.00%
    Total 3,762,863 100%

  3. oiaohm

    Robert Pogson got it in one. Many of the large Linux companies build there own servers in house. Even custom ordered gigabyte boards with on motherboard UPS and other great features.

    IDC is counting full servers sold not parts. To count what Google, Amazon and FaceBook is upto you need to count parts.

    All three order parts by the shipping container full.

    I am not a complete fraud Clarence Moon. Its about time you wake up to this fact.

    Cern and other research made super computers are also not counted due to the fact they are the same.

    We are talking huge sections of the Linux world that don’t register in IDC numbers. IDC does not handle the event of a company ring up a factory in china and having them spit out 100 thousand custom motherboards with integrated UPS feature only for that company. Rackspace a server farm also does the same thing.

    There is at least 20 major computer hardware consumers IDC numbers don’t cover. It also very hard to number these companies because you don’t really know what factories they will be using. Pick and place machine can just as simply make a x86 motherboard as they can make a motherboard for for a dvd player or set-top box.

    Sorting out what was a consumer hardware order and what was hardware going to big high end data centres is kinda impossible.

    IDC numbers are only a part view of the market. It cost 3000 USD to setup a production run of what every board you like. When you need a 100 000 units setup cost is bugger all.

    Cost savings direct ordering from the production plants also do add up. You are talking less than half the cost of the same board retail or general market. There is no sales commission.

    Basically its cheaper for large data centres to build there own then buy from anyone. Also it comes into a waste maths. Upgrading blade cases on site equals less shipping costs.

    Its is also not extra staff cost since the staff have to be there in case of failure or for security anyhow.

    Everything is based on logistics for these super large data centres. Even when Microsoft started trying to build there cloud servers Microsoft found out to be competitive they would have to employ staff to build there own.

    Large data centre with HP/IBM parts selling hosting that is not a HP or IBM data centre does not happen.

    High end build yourself is the rule.

  4. Robert Pogson

    Uh, Clarence, Google is on record as building its own servers. They are not in the business of selling them so what market value will IDC place on that? Same with Amazon and FaceBook.

  5. Clarence Moon

    “IDC for one does not count stuff facebook and google and amazon buys.”

    You are a complete fraud, Mr. Oiaohm. You make up things and deal them out despite the transparancy of you fabrications. You need to get some help.

  6. oiaohm

    IDC for one does not count stuff facebook and google and amazon buys.

    Both of those are doing direct orders to china plants that IDC does not get numbers from.

    There is a huge percentage of the Linux market IDC does not count. Also IDC does not count service contract with Linux distrobutions and Linux Support companies. Every chance as large as what IDC counts again is the real market.

    IDC numbers are basically swiss cheese. Clarence Moon.

    Linux picture is far more complex than most people dream. Like google, facebook and amazon all don’t pay support contract for Linux to any one. Instead have there own in house teams. 20 percent of the time google in house team works on external projects. Facebooks does about 10 to 15 percent and amazon does about 5 percent.

    Basically how do we count the investment of these man hours Clarence Moon. The number of hours invested by companies a year in open source projects is growing.

    Really how to measure Linux income that is the problem. Linux income is part cash and part developer time. Remember MS has to fund all the coders working on core windows.

  7. Clarence Moon

    There seems to be some sort of conflict in and among these numbers, Mr. Pogson. That seems to be a chronic condition around here, I know. IDC is counting, by their taxonomy claims, the cash value of the hardware and operating systems as shipped by the server company factory. The HPC industry figures appear to include the total value of the industry surrounding the use of these computers and consist mainly of application software and the largely in-house research costs associated with the use of these machines than the factory cost of the computers themselves.

    It all gets back to a definition of what is the business at hand. For Microsoft that is selling licenses for its OS products and for its office application products and some associated things such as development products and services. They continue to do quite well at that business.

  8. Robert Pogson

    Clarence Moon wrote, “HPC and web applications are a very minor part of the server market, Mr. Pogson.”

    You wish…

    Big data, by itself is spending $600 million per annum on HPC with 10% p.a. CAGR. Almost anyone doing simulations in R&D will use HPC. The whole market for HPC is estimated to be more than $25 billion per annum. That’s about half the factory revenue for all servers. Considering that 90% of the top 500 use GNU/Linux. There’s a hell of a lot using GNU/Linux beyond the Top 500. IDC is not counting that because people install custom GNU/Linux systems rather than buy them factory pre-installed.

  9. Clarence Moon

    But HPC and web applications are a very minor part of the server market, Mr. Pogson. Furthermore, the Linux server business was growing at rates in excess of 100% per year back then and the predictions were that it would take over the market in short order. Instead Linux has settled into third place, just behind conventional Unix servers.

    “Quarterly revenue of $6.3 billion for Windows servers represented 49.7% of overall quarterly factory revenue.”

    Basically says it all. Microsoft, with 3 times the annual sales increase more in absolute terms than Linux. Hence the gap between them widens with each passing year.

  10. Robert Pogson

    Clarence Moon, in another vain attempt at revising history, wrote, “The same sort of predictions for server OS were being made years ago for Linux, but the trend topped out and began to reverse a couple of years later. It was a flash in the pan, so to speak.”

    GNU/Linux on servers has had ever-increasing growth for a decade or longer. e.g. 90% Top500 HPC systems are GNU/Linux. 60%+ web servers are GNU/Linux.

    IDC reports that server unit shipments are increasing 4% per annum while servers shipping with M$’s stuff increased only 2% in units. GNU/Linux server revenue increased 12% per annum, more than double the rate of growth of servers shipped with that other OS.

    Six times as many EC2 cloud servers run GNU/Linux than that other OS.

    So, GNU/Linux is certainly not a flash in the pan on servers. Only about 10% of the schools where I have worked have a server running that other OS.

  11. Clarence Moon

    Is there any other source of data that shows a correlating trend? The same sort of predictions for server OS were being made years ago for Linux, but the trend topped out and began to reverse a couple of years later. It was a flash in the pan, so to speak.

  12. Flying Toaster

    Correction:

    “And not in maybe 3035, given all data points provided by NetMarketShare.” – > “And not in maybe 3035, given all data points provided by NetMarketShare?”

  13. Flying Toaster

    a rate of growth like that would make GNU/Linux the dominant desktop OS in 3.5 years

    And not in maybe 3035, given all data points provided by NetMarketShare.

    And this is ignoring the way NetMarketShare collects and presents its data, which is crucial in determining if a meaningful conclusion can be drawn.

Leave a comment