Robert Pogson

One man, closing all the windows.

Evolution of IT at a Global Business

  • Mar 25 / 2012
  • 45
technology

Evolution of IT at a Global Business

Years ago Sabre Holdings, a big player in the travel business, used UNIX operating systems on SPARC servers. A few years ago they started phasing out older UNIX servers for GNU/Linux on Intel for huge savings and better performance:
“The move to Linux and Intel has saved Sabre substantial amounts of money, Wiseman says, but it was only possible because he was convinced that it would prove reliable enough to replace the NonStop and Solaris systems. As things turned out, Red Hat’s Linux has proved rock solid, vindicating his position. “In fact we haven’t had a single outage because of Linux,” says Wiseman. “The only outages we have had are from hardware failures, but because the hardware is so cheap we can have redundancy to overcome that, he says.””

Sabre did look at that other OS but decided they did not want the lock-in. Having decided to move to Intel processors, they did not want to have to change software again if they moved to another CPU and GNU/Linux runs on everything…

Recently HP has been their “goto” organization to manage their empire of IT and HP has committed to supporting RedHat on their servers for “mission-critical” applications as more of their customers move to GNU/Linux. Sabre has made a deal for support from HP for six years and $800 million.

Sabre has a lot of good things to say about RedHat GNU/Linux for their global network with 24×7 uptime, three times faster and 90% cheaper than UNIX on SPARC on thousands of servers:

While the use of FLOSS and GNU/Linux on servers for big businesses is easy to understand on price/performance, the same advantages come to businesses who use GNU/Linux on the desktop. I recommend Debian GNU/Linux because it works for me and schools where I have worked but RedHat seems to be the preference of many businesses because of the support and business-oriented attitude of RedHat.

45 Comments

  1. oldman

    “Just like Linux, it is a decrepit and ancient and burdened OS that is fit only for legacy applications.”

    Ahem..

    “Bank of China Core Banking Benchmark revealed IBM System z9® mainframes, working together, generating 9,400 transactions per second, while servicing 380 million customer accounts and approximately 1.3 billion account records. Those results exceed by four times what HP claims in its Temenos Core Banking benchmark. And, System z9 mainframes doing that work required roughly ¼ of the energy to execute”

    “https://www.ibm.com/developerworks/mydeveloperworks/blogs/commonsense/?lang=en_us”

    When loaded with a microcode change that implemented the integrated facility for Linux, an SINGLE system z running zVM with red Hat enterprise Linux as a guest can easily support 1000 virtual Linux Servers at a cost that work out to roughly $500 US per VM.

    Just an FYI ;-)

  2. Clarence Moon

    z/OS is clearly a proprietary IBM mainframe system by this measure

    You will bring a tear to the eye of Mr. Pogson and more sputtering nonsense from Mr. Oiaohm, who are both desperate to “prove” that some 1.7 billion bucks worth of “naked” servers, that they presume are destined to be loaded with some form of Linux (Debian being the local favorite), exist as part of the IDC count. Of course, IDC itself is fairly explicit on the issue, and deems that $1.7 billion to be the mainframe share.

    “IBM’s System z servers running z/OS experienced the second consecutive quarter of positive revenue growth, with 69.1% year-over-year growth in 4Q10 to $1.7 billion. This is the fastest quarterly revenue growth IDC has ever reported for IBM mainframes…”

  3. Clarence Moon

    You keep saying this how many times do I have to pull you up.

    Don’t flatter yourself, Mr. Oiaohm. You haven’t achieved anything beyond showing yourself as a buffoon. Your clumsy constructions and juvenile theories paint you as a complete fool and your lack of real qualifications is transparent to everyone. Why do you persist in embarrassing yourself this way?

  4. aardvark

    I think that anything descended from either System V (AT&T) or BSD (Berkeley, obviously) can lay legitimate claim to being a “Unix.” You don’t actually need a piece of paper to prove your heritage. This group would still include AIX, Solaris, and HP-UX, and probably a few bit-players.

    There’s nothing wrong with claiming to being “inspired by Unix.” There’s nothing wrong with claiming that you “took the best of Unix and improved it,” although tbh you are starting with a pitifully small base.

    This is actually one of the few cases where “counting lines of code” (mostly in the kernel, although I think filesystems and the like are fair game) is a useful metric.

    z/OS is clearly a proprietary IBM mainframe system by this measure. It’s far less of a Unix than is NT/Interix.

  5. ch

    The history of the original Unix is quite tangled, and it seems to end here:
    http://www.sco.com/products/unixware714/
    (Latest release in 2006, so I’m not holding my breath for future development, and anyway – I mean, SCO ? Not saying anything against the original SCO – now Tarantella, but this ex-Caldera outfit is really bad, so it’s no surprise they changed their name to UnXis.)

    The closest surviving relative is probably Solaris.

  6. Robert Pogson

    There isn’t actually a “UNIX” in production as far as I know but there are UNIX operating systems around like AIX, and HP-UX and Solaris. There are hundreds of members of the Open Group, people interested in an open standard for UNIX operating systems and a few who actually release a product. There are lists and UNIX has a short one.

  7. ch

    Mr Pogson,

    another post of mine got eaten here. Please … thanks!

  8. ch

    Mr Ham,

    you sure wrote a lot of stuff in your last comment, but there wasn’t a single fact in it that I haven’t known for years (except that I wasn’t sure whether Interix was Unix-certified). What made you even think for a minute that I wouldn’t know the history of Unix? (I could even point out some small mistakes/simplifications in the graphic you linked to.) However, you have choosen to defend an indefensible position, and it shows.

    “Unix is not a OS. Unix is a certification.”

    That’s so obviously silly that you really have to out-moron yourself (not an easy task) to put it forward. A “Unix certificate” from the OpenGroup means that your OS is _compatible_ with Unix, but it doesn’t turn your OS into something completely different. Or to put it another way: Any definition of Unix that puts OSes as diverse as Z/OS, Windows and Solaris into the same group is _useless_. There is a good reason why the WP entry on Unix starts with “Unix (…) is a multitasking, multi-user computer operating system originally developed in 1969 (…)”. Unix and Unix-like systems share some characteristics that distinguish them from other OSes, and if you wanted to know anything about them, you would find them in ESR’s book. But once more you prefer “believing” and “insulting” over “actually knowing”.

    BTW yes, ESR _is_ a raving lunatic – after all, he is one of the prophets of FLOSS. But he does have a clue about Unix, and therefore his book is valuable, while your ranting is not.

  9. Robert Pogson

    NetApplications’ numbers also come from server logs but I have proven they must be counting them only during business hours or connections from corporate domains. That way, Google’s GNU/Linux numbers blow things up. That other OS is getting a free ride from all the locked-in businesses whose instances are being counted while non-business use of GNU/Linux is not. There’s no other explanation for the effect of a single business on the whole number for the US. Excluding San Francisco cuts the US number down by a factor of 3.

  10. oiaohm

    Clarence Moon

    “IDC’s numbers are counts of server unit shipments worldwide”

    You keep saying this how many times do I have to pull you up. IDC counts revenue that is the only thing they count all there numbers are based off that.

    What are IDC unit shipments from IDC. “unit shipment estimates” That is in IDC own words. They take revenue numbers divide by some magic number of guessed average price and get Units. Basically IDC never counts number of units.

    Lot of makers number of units produced are internal figures only. Why if you know how many units a maker is turning over you can work out as a supply if that company is sourcing from another supplier as well for the same part.

    If you don’t understand what IDC does you cannot use there numbers. IDC is revenue counting so has to be taken with a fairly large grain of salt that is Unit sales estimates could be way out. In fact be multi times out in either direction for any OS in there counts.

    IDC unit numbers are not worth a cracker never have been worth anything. Problem Clarence Moon you keep on thinking they are worth something.

    Percentage of Income is only useful to see roughly if a party is growing on reducing in market share not how much they have.

    If you want to work out unit counts you need to work out someway to get makers to give them to you or do ground surveys.

    Web numbers are fairly well screwed up as well.

  11. Clarence Moon

    If IDC’s numbers are biased that way because of 10K users of GNU/Linux in business,

    Maybe they would if IDC’s numbers were biased that way, but they are not. IDC’s numbers are counts of server unit shipments worldwide and breakdowns of financial revenues for server manufacturers. It has nothing to do with San Francisco, Google, or Timbuktu. IDC has nothing to do with Net Applications. I think you are confusing the two.

  12. aardvark

    Mr Oiaohm:

    A rare Unix system without a blood line back to “Unics” (incidentally, Linux has no blood line back to *nix”) is not, in fact, a Unix system. You might as well argue — and I have no doubt you will, should the mood take you — that a dachsund has the same blood line as a pure-bred Arabian Stallion.

    In a sense, that is true. In a truly worthless sense, that only you would find convincing.

    You really have to be a raving nutter to insist that z/OS “is a Unix.” The bloodline runs from S/360 to S/370 to S/380 to OS/390. Possibly it takes on AS/400 and System/1 on the way; I couldn’t say. At no point whatsoever has it been polluted by the wretched germinal waste of Unix.

    Just like Linux, it is a decrepit and ancient and burdened OS that is fit only for legacy applications. But that’s about as much as the two have in common.

    “So yes some of my hate of Microsoft is they were given a almost perfect OS design by Dave Cuttler then they allowed there coders to completely ruin it.”

    VMS was far from “almost perfect.” Like all Multics children, it was a work in progress. There’s a very good case to be made that Cutler’s team actually achieved this by NT 4.0, and to be completely objective here, no matter how I wish things had turned out differently, Windows 7 is pretty much what Cutler and the VMS team would have aimed at in the first place.

    I live in hope that ignorance will spawn fascinating insight, however. Perhaps you would care to share with us all the precise details (even the vague handwavy links) of how Microsoft allowed their coders to ruin Mr Cutler’s dream?

    Every time I believe I have seen you plough the depths of determined ignorance, Mr Oiaohm, you continue to surprise me. Is there no limit at all to how much of a cretin you can be?

  13. oiaohm

    ch z/OS has a Unix certification not just a Plan Posix one.

    oiaohm
    “If NT design worked as designed you should be able to load up isolated win32/win64 environments side by side with each other on the one NT kernel.”
    ch
    -Why would I want to do that? The Win64 personality happily runs my Win32 and Win64 stuff alongside each other, so why run them in different environments?-

    You did not read. Its like two Win64 environments running next to each other and not being able to see each other. Perfectly sand-boxed from each other. NT was designed to pull this off before ruined.

    Cgroups on Linux, Jails on BSD based and Zones on Solaris are all related ideas. These are having to be hacked into the ABI’s. Windows had this stuff designed in at the start.

    Then developers at Microsoft shot it in the way they allowed subsystems to bind up. Most people don’t know windows kernel has a internal thread table and a process representation thread table. Reason for this existing was to allow like two win32 or two win64 environments to run side by side on the same kernel without fighting.

    Yes the NT design was head of its time by many years. So far that the other coders working at Microsoft failed to see the design and ruined it majorally.

    So yes some of my hate of Microsoft is they were given a almost perfect OS design by Dave Cuttler then they allowed there coders to completely ruin it.

    Ch also please look back Interix was a certified Unix95 when on Windows.

    Unix is not a OS. Unix is a certification. Any OS can be certified Unix as long as it passes the tests. There is no requirement for blood line or anything else.

    This is the mistake lots of people do Unix OS that is not real. z/OS is a certified Unix just like AIX Solaris and many other things are.

    Ch basically find Unix. http://upload.wikimedia.org/wikipedia/commons/7/77/Unix_history-simple.svg

    It don’t exist as a OS. Relatives of Unics exist.

    The closet you will find is Unixtss.

    Nothing in opengroup rules say that a Unix has to be a linked to the Unics blood line. This is why Linux in the past has been a Unix. The mistake that people who don’t read opengroup allows them to rule z/OS out.

    Even so if you dig deep enough into the Linux kernel you will find some Unix files that have crossed into Linux from BSD in drivers.

    Yes there are drivers that are mostly shared between the BSD line and the Linux line. So the graphic shows them as independent when Linux should be linked onto the side of BSD.

    What ESR says does not match what Opengroup says. Opengroup are who say what Unix is. So unless you are a define of Unix from the Opengroup its bogus ch.

    If people learn to read source documents a lot of problem would be solved instead of reading third party invalid hear say like ESR talking about Unix.

    Eric S. Raymond as not updated his FAQ since 2003. Most likely never will again. He did get into trouble over it. The history in that faq is also incorrect by Opengroup documents. Opengroup is the owner of the Unix trademark so defines how it is used. Eric S. Raymond has completely zero standing on what is Unix.

    Really ch quoting Eric S. Raymond shows how big of a idiot you are. Laughing stock or what. Quoting someone who has no legal right to say what is and what is not Unix as some kind of define. Really learn to do research properly ch and check out if the document you are referring is anything more than toilet paper at best if printed.

    http://en.wikipedia.org/wiki/Single_UNIX_Specification

    Sorry reality is reality z/OS is a Unix. Its one of the rare Unix systems without a blood line back to Unics.

  14. Robert Pogson

    California is a block of 37 million people. If NetApplication’s numbers are biased that way because of 10K users of GNU/Linux in business, imagine how biased the numbers for the world are.

    UPDATE I mistakenly wrote “IDC” when I should have written “NetApplications” here.

  15. Clarence Moon

    If you really believe IDC then you must believe GNU/Linux has 13% share in California

    Can you find a cite where IDC says that is the case? Net Applications has some defect in their counting scheme or else in their data logging scheme that creates pockets of obvious nonsense. The count hits on client websites. You have demonstrated that their methods are not effective, at least when applied to narrow segments of their data. Overall, there is no doubt that their reporting of total hits throughout their coverage area is accurate, but you can question whether or not their application area is representative of the actual installed base.

    I doubt that it is, simply because, as I have previously posted, they do not cover the most popular sites, namely Google, Facebook, and Twitter. Nor do their statistice reflect what happens at Yahoo, MSN, CNN, or other major gateway sites. It does not address web services consumed by apps on phones or tablets either. Those are glaring omissions of things that are important to market strategies.

    What they do show is that less than 1% of the people visiting their client’s sites use desktop Linux. Who those people are can be argued, but there is nothing to be gained by knowing the answer.

  16. ch

    Mr Pogson, my last two posts were swallowed by the spam filter, so could you please … ? Thanks!

  17. ch

    “Sorry z/OS is a good example that UNIX is a grey define.”

    It is a good example why “able to run Unix applications” is not the same as “being Unix”. The various POSIX and Unix certificates essentially only care about the former. ESR pretty well describes what “being Unix” is about here:
    www dot faqs dot org slash docs slash artu
    (Sorry but it seems the spam filter gets extra hungry if I put links in my post.)

    “OS X is also currently a certified UNIX.”

    Of course, technically it IS a Unix. But Apple was clever enough from hiding most of the unixy stuff from users. To quote Joel Spolsky again:

    OS X is the proof: Apple finally created Unix for Aunt Marge, but only because the engineers and managers at Apple were firmly of the end-user culture (…). They rejected the Unix culture’s fundamental norm of programmer-centricity. They even renamed core directories — heretical! — to use common English words like “applications” and “library” instead of “bin” and “lib.”
    www dot joelonsoftware dot com slash articles slash Biculturalism dot html

    (If ESR’s book is too long for you, JS’ article brings you the highlights in regard to “what is Unix”.)

    “If NT design worked as designed you should be able to load up isolated win32/win64 environments side by side with each other on the one NT kernel.”

    Why would I want to do that? The Win64 personality happily runs my Win32 and Win64 stuff alongside each other, so why run them in different environments?

    “Also you would be able to load up subsystems to emulate anything.”

    The OS/2 subsystem is dead, but I don’t see many people mourning that. The Posix subsystem is still around – btw, it’s now officially called “Subsystem for UNIX-based Applications”. So what are we missing? Do you really need a Z/OS subsystem on Windows? ;-)

  18. ch

    “M$ changed things arbitrarily in its OS to make moving back to UNIX a jarring experience.”

    No, they didn’t. MS hired Dave Cuttler, so it’s not really surprising that the NT kernel is “influenced” by VMS (not the worst pedigree). The network stuff came basically from LAN Manager (at the time a network system with a lot of industry support behind it: IBM, 3COM, HP, AT&T, SCO) and the base for the GUI from Windows (although they probably didn’t reuse much actual code). Of course, the team looked around at what else there is out there, and it would be surprising if they hadn’t found _something_ useful to learn in Unix, too – and they probably learned what _not_ to take from Unix, as well. The result was a completely new core OS with personalities compatible with Windows, OS/2 1.x and Posix, respectively.

    The problem with POSIX is that it’s so limited POSIX alone won’t take you anywhere useful. So the early POSIX personality on NT was useless but still standard-compliant. Later, MS bought Interix which AFAIK is an actually useful POSIX-compliant Unix-like personality, so you could turn your XP/Vista/7 into a unixoid system. Of course, this doesn’t mean that “Windows is Unix”, just like z/OS isn’t Unix just because it is certified as Unix-compatible.

    http://technet.microsoft.com/library/bb463212

  19. Robert Pogson

    Clarence Moon wrote, “If you want to dispute this, you should be able to show where a statistic from the past was found to be in error and no such thing exists. Certainly if it did, there would be a lot of publicity given to it.”

    My article on the subject was rejected by Slashdot…

    Really, look at the time series at Mountain View, California – Penguin Heaven and
    http://mrpogson.com/2012/02/05/whats-happening-with-floss-in-usa/

    10K people made a huge increase in NetApplications’ data for all of California.

    If you really believe NetApplications then you must believe GNU/Linux has 13% share in California when nothing rises above 1% except the area around San Francisco. Of course, that’s not believable. There’s no possible mechanism for it except NetApplications overcounts businesses and Google is a business.

  20. oiaohm

    Robert Pogson
    http://www.phil.muni.cz/~letty/misc/open_linux.html

    There were about another 7 others.

    aardvark
    “I believe the Linux Standard Base and Linux SE are Posix-certified, although I wouldn’t swear to it.”

    That is wrong Linux Standard Base is a response to OpenGroup bills for Unix and Posix certifation.

    Even that Linux Standard Base contains all the test for Posix and Unix certification since money has not be paid to opengroup it is not Posix certified or Unix certified.

    Linux SE is something else completely.

    When Linux world did make Linux Standard Base this all the certified Unix Linux systems got deleted from the Opengroup register. So hiding the fact that Linux was registered at a particular point in time.

    aardvark recertification with Opengroup use to require paying a fee each time. Its the Linux Standard base that forced Opengroup to go to a yearly fee.

    Lot of ways the opengroup has acted like a spoiled brat.

    DoD accept Linux Standard Base certification as the same as Unix or Posix certification in acquirement.

  21. Clarence Moon

    sigh… One number being wrong does not make them all wrong.

    Aside from the fact that it truly does not matter one whit to me or you whether the numbers are right or wrong, it seems to me that you are being somewhat disingenuous to suggest that the numbers that show something that you agree with are correct and the numbers that show something that you do not want to be the case are wrong. Neither one of us are going to use the information for any useful purpose.

    IDC sells this information to people who make business decisions based on its accuracy. These people have been buying this information for decades now and continue to buy it which suggests that they find it both useful and accurate. If you want to dispute this, you should be able to show where a statistic from the past was found to be in error and no such thing exists. Certainly if it did, there would be a lot of publicity given to it.

  22. aardvark

    Mr Oiaohm:

    “Its evil because all updates have to be recertified that they have not broken conformance with the Unix standard.”

    I gotta admit, that is evil incarnate.

    Shouldn’t the people involved with the standard just take your word for it? It seems such a shame that they expect a degree of conformance, or even sanity.

    What has the world of standards come to?

  23. aardvark

    I believe the Linux Standard Base and Linux SE are Posix-certified, although I wouldn’t swear to it.

    No distros that I am aware of. Posix is a weapon exclusively used to hammer NT with, even though NT was expressly developed to allow “personalities” (ie different OS superstrates) and therefore Posix compliance. Which it was awarded.

    Presumably even the DoD has given up on Posix compliance, since they seem quite happy to use various Linux distros. But then, it’s a twenty year old self-defined standard (I’m using self-defined loosely here: the actual history, to do with the original Unix Wars in the 1990s, is too hideous to go through at this point).

    So who cares?

  24. Robert Pogson

    Yes, it’s the release early and often aspect of GNU/Linux which makes certification unreasonable for most distros. The ones with long term support and commercial backing could do it, I suppose.

    What distro was ever certified? RedHat? Suse?

  25. oiaohm

    ch and Robert Pogson you are both wrong.

    Linux at particular points in history has been a certified Unix. It is an highly costly process to pass certification as a Unix. Its evil because all updates have to be recertified that they have not broken conformance with the Unix standard.

    Shockily Windows NT has also been certified Unix at particular times as well.INTERIX at one point it did pass and MS was allowed to use UNIX trademark. Yes NT was not just Posix certified but a UNIX as well.

    Being a Certified Unix OS does not forbid the OS from having its own ABI as well.

    Sorry z/OS is a good example that UNIX is a grey define.

    OS X is also currently a certified UNIX. So every OS X user is a Unix user by open group standards.

    ch NT supports directly split of either / or \ Yes the / is for posix and unix support. NTFS even include a sub section for posix filenames a long side windows filenames.

    Yes OS X server are Unix servers as well.

    Really from my point of view Unix marker it too vague to really tell you anything.

    And you are wrong I don’t want NT to be Unix like. I would love to see NT implemented how it was properly design. With its subsystem design working properly allowing isolated environments.

    NT design is super good. Far better than Linux design at current day. Microsoft stuffed it with there implementation. So why would I want a superior design to be made Unix like.

    If NT design worked as designed you should be able to load up isolated win32/win64 environments side by side with each other on the one NT kernel. Also you would be able to load up subsystems to emulate anything.

    That is the way NT is designed. Way NT is implemented is a ruined mess.

  26. Robert Pogson

    see http://aplawrence.com/Unixart/gates_quote.html

    M$ changed things arbitrarily in its OS to make moving back to UNIX a jarring experience.

    No one can fix that other OS except M$ and they won’t talk to little old me. Even a fresh install of that other OS soon results in problems in my experience. That’s why re-installing it is a growth industry. I have rarely needed to re-install GNU/Linux since I moved to Debian with its brilliant APT packaging system.

  27. ch

    Sorry Mr Pogson, but could you please “resurrect” my last post? Thanks!

  28. ch

    “M$ is the only one who has claimed NT was UNIX-like.”

    Actually, MS never claimed that. But NT shipped with a POSIX-compliant “personality” in addition to it’s Windows and OS/2 1.x “personalities”. Today, you can install the “Services for Unix” to get an improved POSIX personality. (The original POSIX personality was pretty much useless but still allowed MS to sell NT as POSIX-compliant.)

    “I have seen nothing in NT that resembles UNIX at all except FDISK perhaps.”

    That’s the wrong place to look at – better try the system calls for opening / reading / etc. files.

    “Where is the root user?”

    It’s called “System” – and the fact that you don’t know that explains a lot about your problems with Windows.

    “Whatever happened to forward slashes?”

    As every user of the former OS standard CP/M knows, forward slashes indicate command parameters, not directories.

    “It was crashing weekly on my wife’s PC in the early days.”

    And you couldn’t fix that? Why didn’t you ask someone who actually knows Windows?

    “GNU/Linux is definitely UNIX-like even if it was never certified.”

    Of course, but according to Mr Ham’s twisted logic NT should be more Unix-like.

  29. Robert Pogson

    M$ is the only one who has claimed NT was UNIX-like. I have seen nothing in NT that resembles UNIX at all except FDISK perhaps. What has “the registry” got to do with UNIX? I wonder. Where is the root user? Was there ever an uptime command? That must have been embarrassing. Whatever happened to forward slashes? I’ve only used the original NT at one employer. We always feared to print anything. It froze about 10% of the time. XP, is supposedly NT-based but it surely wasn’t UNIX-like in its uptime when new. It was crashing weekly on my wife’s PC in the early days. Then it increased/improved to monthly and later a bit better than that. Last year, I just got fed up with it freezing and after verifying that the little woman could do everything she needed using GNU/Linux, I paved it over. She has some problems with one PC but it’s a flaky notebook that was dropped at the airport and has only 512 MB RAM. Chrome web browser and a lot of tabs open seems to run it short of RAM often. XP would be a dog on that machine but it flies normally. GNU/Linux is definitely UNIX-like even if it was never certified.

  30. Robert Pogson

    ” The Open Group certifies z/OS as a compliant UNIX operating system — with UNIX/Linux-style hierarchical HFS and zFS file systems.”

    see http://en.wikipedia.org/wiki/Z/OS

    Interesting. I didn’t know that. Thanks, oiaohm. I had assumed it was more of the same thing as System 360/370 etc. that I knew in the 1960s and 1970s. Those were definitely not UNIX.

  31. Robert Pogson

    sigh… One number being wrong does not make them all wrong. The trend in a set of wrong numbers can be correct even if all the numbers are wrong. e.g. y=mx+b still has the same m even if b is replaced by b+4.

    I have been number-crunching and analyzing data since the 1960s. Look up random and systematic errors. Look up blunders, typos, etc. I have proven IDC’s numbers are wrong but they still measure something of value, usage in business, as far as I can tell. A business like Google using GNU/Linux shows up loud and clear in their data. A school system using GNU/Linux does not. You figure out why. I can only guess their methods and motives.

  32. ch

    “z/OS is a Unix”

    Thanks for lighting up my day! Oh, and since WinNT was certified POSIX-compliant while Linux (at least at the time) wasn’t, that surely means that WinNT is more Unix-like than Linux is, right?

    No, z/OS is NOT Unix, not in any way, shape or form. It just has an additional UNIX API next to it’s original API that allows customers to run Unix apps on Z/OS.

  33. Clarence Moon

    That is illogical

    Not at all, Mr. Pogson. It would be illogical to, as you are doing, accept one set of numbers from IDC, since they allow you to make some point, and reject others because they do not support your pre-conceived notions.

    It is logical to think that IDC, whom you trust to count dollars, can also count units. It is logical to think that OS products that are applied to very high cost units such as HPCs or large servers will have lower unit volume for their revenue contributions than the OS that is mostly applied to lower cost units.

    Just listing a configuration on a web page does not imply that many are being sold. The fact that the IDC report does not bother to even provide a category for such machines in its abstract summary argues that the category is unimportant to IDC’s customers which would not be the case if it were a major part of the market being reported.

  34. oiaohm

    IDC does not count the parts market either. Lot do end up in the Linux world as well.

    z/OS is counted in UNIX. Clarence Moon you are a moron with the numbers as normal.

    $1.7 billion unaccounted does not include Z/OS never has. Pulling fake straws as normal because you don’t know a single thing about Operating systems.

    “The Open Group certifies z/OS as a compliant UNIX operating system.” Yes z/OS is a certified Unix.

    IBM makes two Unix OS’s one being AIX and the other being z/OS. Anything with either installed is counted in Unix section of IDC numbers.

    $3.4-1.7=1.7. Yep z/OS is half if the remaining Unix sold. When you add in AIX IBM is the dominate Unix maker left without question.

    Come on Clarence Moon you take about me making a fool out of myself. This is just stupidity on your part not knowing that z/OS is a Unix. I guess you did not know that IBM is the dominate Unix maker either.

  35. Robert Pogson

    Clarence Moon wrote, “If you accept one set of numbers from IDC to prove your point, you must logically accept all the numbers from IDC or else cast doubt on your own assertions.”

    That is illogical. You provided the unit share % from IDC. You defend it any way you can. I don’t accept that number. That does not reflect at all on the revenue share numbers.

    The revenue share shows nothing left for naked servers. The revenue share does not account for virtual machines, either. M$’s unit share of real machines could be small still and a server could be running both operating systems.

    Dell, at least must be selling a significant number of naked machines or they would not list them.

  36. Clarence Moon

    How is it possible to get that much revenue (45%) from servers while getting 75% of servers?

    You have to ask yourself, just what is your purpose in all this, Mr. Pogson. Are you trying to say that IDC is wrong? If you accept one set of numbers from IDC to prove your point, you must logically accept all the numbers from IDC or else cast doubt on your own assertions.

    The answer is what I know I have posted before in this forum, namely that a great deal of the Linux server business is on big machines. You yourself note that the million plus dollar HPC machines are mostly Linux. Also, a large number of big servers from HP or Sun/Oracle or IBM that used to be Unix are now shipping with Linux. So a lot of the Linux server revenue is concentrated in these big, expensive units. Microsoft Windows, on the other hand, is concentrated in the low end of things. There are some big iron Windows sales, certainly, but they are few compared to Linux and Unix.

    So with almost half of the revenue overall and very little of the high end business, Microsoft makes it up in quantity of low end servers and that gives them a high unit volume.

    That means IDC is not counting naked servers at all and their numbers are bogus for share of units shipped.

    I don’t follow that flurry of numbers, Mr. Posgon, but I would say that it may very well be that what IDC is counting is what they say they are counting and that is revenue from sales of servers world-wide. There may not be any measurable quantity of “naked” servers. My own experience with our customers is that they often buy our products as part of a purchase from IBM, Dell, or HP that includes the OS as well as our stuff. I don’t know for sure just what happens with non-Windows servers, but I know that a lot of the Linux and Unix product goes out the same channels. If someone were buying a lot of servers and installing some free version of Linux themselves, they would, like as not, not buy our product anyway. It costs more than the OS.

  37. Robert Pogson

    That means IDC is not counting naked servers at all and their numbers are bogus for share of units shipped.

    IDC says 2.2 million servers were shipped in Q4 2011. For a revenue of $5.8 billion that means an average or $5.8 billion/(2.2 million X 75%) = $3500 for the average price of a server with M$’s stuff. That means M$ was getting $0. Not likely. Their revenue for a quarter is $billions. Again 75% is very unrealistic. M$’s revenue for servers divided by 75% of servers comes to $4.7billion /(2.2 million X 75%) = $2848 per server, far less than their actual take including CALs.

  38. Robert Pogson

    Clarence Moon wrote, “HPC is but a handful of units annually, Mr. Pogson, surely you must understand that.”

    Uh, no. What HPC lacks in units it makes up for in rack-space.
    “March 20, 2012 – Worldwide factory revenue for the high performance computing (HPC) technical server market increased by 8.4% in full-year 2011 to reach a record $10.3 billion, up from $9.5 billion in 2010, according to the newly released International Data Corporation “

    see IDC

    HPC is huge. It’s use is obvious in CAD/CAM and such but also any huge business needs some HPC just to figure things out. These are the calculators of big data.

    The 75% number is inconsistent with all the data I have seen. How is it possible to get that much revenue (45%) from servers while getting 75% of servers? That would imply lowered prices or lower-end servers. That just doesn’t work. We know M$ causes higher prices.

  39. Clarence Moon

    so it is quite unlikely that M$ can hold that large a share in unit volume

    HPC is but a handful of units annually, Mr. Pogson, surely you must understand that. Similarly, web servers are but a minor share of the overall market for servers. You use IDC’s numbers for most of your reply, so you must accept their numbers for unit volume. We have been over them more than once here in the recent past and IDC is reporting some 75% of unit volume for Windows servers in one quarterly report as recently as a year ago.

    “Microsoft Windows server demand also continued to show strong demand as Windows-based hardware revenue increased 10.1% year-over-year. Quarterly revenue of $5.8 billion for Windows servers represented 48.5% of overall quarterly factory revenue and 75.2% of all quarterly server shipments.”

    That leaves $14.2-($6.5+$3.4+$2.6) = $1.7 billion unaccounted, probably naked servers.

    Can you say “mainframe“?

    “IBM’s System z servers running z/OS experienced the second consecutive quarter of positive revenue growth, with 69.1% year-over-year growth in 4Q10 to $1.7 billion.”

  40. aardvark

    As always, I trust your figures implicitly, Mr Pogson. For the “thickies” amongst your readers, let me add them up:

    Unix: $3.4 billion = $3.4 billion
    Linux: $2.6 billion + $1.7 billion = $4.4 billion
    Windows: $6.5 billion = $6.5 billion

    The more astute amongst the thickies will notice a small mathematical discrepancy: I have rounded up, in the case of Linux, because the figures after the decimal point are missing.

    It’s not easy to suggest that Linux is ahead of Windows on servers, given these figures. Perhaps, if Linux were to cannibalize 100% of the Unix server market, it might just about edge it (to be fair, it would be ahead by roughly 25%), but do you really see that as an achievable objective?

    And isn’t it a bit ironic to be celebrating what is effectively the death of Sun, HP-UX, and AIX in the marketplace, when what you are apparently wishing for is the death of M$?

  41. Robert Pogson

    Clarence Moon, referring to M$, wrote, “Windows servers account for about half of all server revenues annually. They account for an even higher share of server unit volume.”

    According to IDC about 8.3 million server units shipped in 2011. GNU/Linux blows M$ away in units on HPC (91.4% of supercomputers and 99.9%+ of gigaFLOPs) and web (60%+ while M$ is ~15% and decreasing) so it is quite unlikely that M$ can hold that large a share in unit volume.

    IDC reports that quarterly server factory revenue was $14.2 billion, GNU/Linux revenue was $2.6billion, Windows revenue was $6.5billion, and UNIX was $3.4 billion. That leaves $14.2-($6.5+$3.4+$2.6) = $1.7 billion unaccounted, probably naked servers. In my estimation, a lot of those naked servers got GNU/Linux and the typical server is around twice as much for hardware+software for that other OS so the revenue share is far higher than unit share for M$. I think GNU/Linux is ahead of that other OS in unit share.

    e.g. Dell PowerEdge R515
    $4400 naked. add $3000 for Server 2008 Enterprise with only 10 CALs… With GNU/Linux there is no licensing limit on access.

  42. aardvark

    $800 million is hardly pocket change, is it? I doubt that the difference between RHEL licensing and support on the one hand, and M$ licensing and support on the other, would make much of a dent in that.

    SABRE has about forty years of development, database infrastructure, networking infrastructure, and so on to protect. Which choice do you think they would make, Mr Pogson? Throw it all away and start from scratch on Windows Servers? Or port the whole raft from Unix to Linux?

    It’s something of a no-brainer, really, but if you want to make it out as a win for Linux on technical merit, be our guest.

  43. Clarence Moon

    So what is your point here, Mr. Pogson? It is well known that Unix users have migrated to Linux in many areas where there was no justification for a branded Unix OS that cost a lot more and offered no functional or performance advantage over the work-alike Linux versions from Red Hat or Novell. IBM started this movement more than a decade ago when its AIX products were languishing in the face of Solaris and HP-UX. They shot a hole in the price umbrella that existed for Unix servers and the industry has never recovered.

    Sun is not even an independent company anymore, eaten up by Oracle in an ironic turnaround from their mutual beginnings.

    During that same period of time when Linux was devastating Unix business, Microsoft, starting at zero, has grown over the years to a 78% share of revenues attributed directly to the OS component of a server sale and Windows servers account for about half of all server revenues annually. They account for an even higher share of server unit volume.

    You can try to paint that as some sort of loss for Microsoft, but you are not going to get a very loud “Amen!” from even your own choir. Meanwhile, Microsoft’s server division sales grow and grow, setting new revenue records quarterly and annually.

Leave a comment