World's largest IT company

Discussion in 'NZ Computing' started by Lawrence D'Oliveiro, Dec 17, 2006.

  1. I always thought IBM's mainframe business was so large that it put the
    company comfortably ahead of anybody else in the IT business. But not so,
    according to this <http://www.theregister.co.uk/2006/12/17/hp_number_one/>.
    Looks like HP's acquisitions have now made it rival IBM in size. And yet it
    doesn't sell any mainframes.

    So doesn't that mean that mainframes aren't such a huge business any more?
    Some have been calling them "dinosaurs" for over a decade, while others
    kept pointing out how big a business they were. Looks like that era is
    over, and they really are becoming dinosaurs now.
     
    Lawrence D'Oliveiro, Dec 17, 2006
    #1
    1. Advertising

  2. Lawrence D'Oliveiro

    whome Guest

    "Lawrence D'Oliveiro" <_zealand> wrote in message
    news:em4jdh$6d0$...
    >I always thought IBM's mainframe business was so large that it put the
    > company comfortably ahead of anybody else in the IT business. But not so,
    > according to this
    > <http://www.theregister.co.uk/2006/12/17/hp_number_one/>.
    > Looks like HP's acquisitions have now made it rival IBM in size. And yet
    > it
    > doesn't sell any mainframes.
    >
    > So doesn't that mean that mainframes aren't such a huge business any more?
    > Some have been calling them "dinosaurs" for over a decade, while others
    > kept pointing out how big a business they were. Looks like that era is
    > over, and they really are becoming dinosaurs now.


    I don't know. Can a PC reliably run a utility billing and provisioning
    application with over 3 million customers?

    Mainframes will never be dead.
     
    whome, Dec 17, 2006
    #2
    1. Advertising

  3. Lawrence D'Oliveiro

    Chris Lim Guest

    whome wrote:
    > I don't know. Can a PC reliably run a utility billing and provisioning
    > application with over 3 million customers?


    I don't see why not. Is this a limitation you're aware of?
     
    Chris Lim, Dec 18, 2006
    #3
  4. On Mon, 18 Dec 2006 12:22:42 +1300, Lawrence D'Oliveiro wrote:

    > So doesn't that mean that mainframes aren't such a huge business any more?
    > Some have been calling them "dinosaurs" for over a decade, while others
    > kept pointing out how big a business they were. Looks like that era is
    > over, and they really are becoming dinosaurs now.


    Mainframes are the way to go for large, highly-reliable, mission-critical
    systems that require multiple applications running simultaneously with
    quick and simultaneous access to essential data by thousands of users.

    There still are some things that are so large and so mission-critical that
    a windows box cannot cope with them.

    It all depends on the degree of redundancy and reliability a busness
    requires, and whether or not it wants one Mainframe, or multiple
    individual Unix boxes.


    Aquilegia Alyssum

    --
    "The only way Vista client and Longhorn server would make sense
    would be if [the] company was doing a 'forklift upgrade' on its
    entire client-server infrastructure."
     
    Aquilegia Alyssum, Dec 18, 2006
    #4
  5. In message <>, whome wrote:

    > Can a PC reliably run a utility billing and provisioning
    > application with over 3 million customers?


    A network of PCs could. If you have a large enough collection of them,
    running a suitably robust clustered computing system, who cares if 5% are
    dead at any given moment. Given the economies of scale in the production of
    PCs, the system could indeed be cheaper than a mainframe.
     
    Lawrence D'Oliveiro, Dec 18, 2006
    #5
  6. Lawrence D'Oliveiro

    whome Guest

    "Chris Lim" <> wrote in message
    news:...
    > whome wrote:
    >> I don't know. Can a PC reliably run a utility billing and provisioning
    >> application with over 3 million customers?

    >
    > I don't see why not. Is this a limitation you're aware of?
    >


    PC's may be 'somewhat' capable. However, Mainframes are industrial strength
    machines which have evolved over many years to provide massive transactional
    power with exceptionally high reliability.

    Would you run your core business applications on a somewhat capable system,
    or a most capable system?

    I know what I'd choose, and I'd be in business while you are going bankrupt
    trying to debug application contention and comms problems .
     
    whome, Dec 18, 2006
    #6
  7. Lawrence D'Oliveiro

    whome Guest

    "Aquilegia Alyssum" <> wrote in message
    news:p...
    > On Mon, 18 Dec 2006 12:22:42 +1300, Lawrence D'Oliveiro wrote:
    >
    >> So doesn't that mean that mainframes aren't such a huge business any
    >> more?
    >> Some have been calling them "dinosaurs" for over a decade, while others
    >> kept pointing out how big a business they were. Looks like that era is
    >> over, and they really are becoming dinosaurs now.

    >
    > Mainframes are the way to go for large, highly-reliable, mission-critical
    > systems that require multiple applications running simultaneously with
    > quick and simultaneous access to essential data by thousands of users.
    >
    > There still are some things that are so large and so mission-critical that
    > a windows box cannot cope with them.
    >
    > It all depends on the degree of redundancy and reliability a busness
    > requires, and whether or not it wants one Mainframe, or multiple
    > individual Unix boxes.
    >
    >
    > Aquilegia Alyssum
    >
    > --
    > "The only way Vista client and Longhorn server would make sense
    > would be if [the] company was doing a 'forklift upgrade' on its
    > entire client-server infrastructure."
    >


    Of course you are correct. But, some people here think pc's have no limits
    so long as you can put them in parallel. Bugger that for my core system.
     
    whome, Dec 18, 2006
    #7
  8. Lawrence D'Oliveiro

    frederick Guest

    Lawrence D'Oliveiro wrote:
    > I always thought IBM's mainframe business was so large that it put the
    > company comfortably ahead of anybody else in the IT business. But not so,
    > according to this <http://www.theregister.co.uk/2006/12/17/hp_number_one/>.
    > Looks like HP's acquisitions have now made it rival IBM in size. And yet it
    > doesn't sell any mainframes.
    >

    <snip>
    80% of HP profit last year came from sales of ink and toner.
    For a US$100 billion t/o company, perhaps they should be considered
    first as the world # 1 ink maker, with subsidiaries in the computer and
    peripherals businesses.
     
    frederick, Dec 18, 2006
    #8
  9. Lawrence D'Oliveiro

    Shane Guest

    Lawrence D'Oliveiro wrote:

    > In message <>, whome wrote:
    >
    >> Can a PC reliably run a utility billing and provisioning
    >> application with over 3 million customers?

    >
    > A network of PCs could. If you have a large enough collection of them,
    > running a suitably robust clustered computing system, who cares if 5% are
    > dead at any given moment. Given the economies of scale in the production
    > of PCs, the system could indeed be cheaper than a mainframe.



    Whats google running on these days?

    --
    Bender: I can't keep running people over. I'm not famous enough to get away
    with it.

    blog: http://shanes.dyndns.org
     
    Shane, Dec 18, 2006
    #9
  10. Lawrence D'Oliveiro

    whome Guest

    "Lawrence D'Oliveiro" <_zealand> wrote in message
    news:em4qap$iei$...
    > In message <>, whome wrote:
    >
    >> Can a PC reliably run a utility billing and provisioning
    >> application with over 3 million customers?

    >
    > A network of PCs could. If you have a large enough collection of them,
    > running a suitably robust clustered computing system, who cares if 5% are
    > dead at any given moment. Given the economies of scale in the production
    > of
    > PCs, the system could indeed be cheaper than a mainframe.
     
    whome, Dec 18, 2006
    #10
  11. Lawrence D'Oliveiro

    Dave Taylor Guest

    Lawrence D'Oliveiro <_zealand> wrote in
    news:em55t5$6i0$:

    > About 450,000 servers
    > <http://cbcg.net/talks/googleinternals/index.html>. The servers are
    > cheap x86 boxes, except in one regard--they don't skimp on the power
    > supply. That's because a good-quality supply is more efficient, which
    > reduces power bills.
    >


    For more on the GFS, (Google File System)
    go here, and just search for Google, lots of interesting in depth talks on
    how it really works from people who design and work with it:
    Check out the CSE:
    http://www.researchchannel.org/search/sitesearch.aspx
    http://www.researchchannel.org/prog/displayevent.aspx?rID=4188&fID=345

    There at least two others in the CSE that have interesting info about how
    Google runs its clusters.

    --
    Ciao, Dave
     
    Dave Taylor, Dec 18, 2006
    #11
  12. Lawrence D'Oliveiro

    steve Guest

    Lawrence D'Oliveiro wrote:

    > So doesn't that mean that mainframes aren't such a huge business any more?
    > Some have been calling them "dinosaurs" for over a decade, while others
    > kept pointing out how big a business they were. Looks like that era is
    > over, and they really are becoming dinosaurs now.


    Mainframes now sit quietly under tables in normal rooms.

    Mainframes have a firm niche in recent years as "super servers".

    IBM sells mainframes that can run many MVS images or literally thousands
    of Linux images - each a server in its own right for some user community
    it serves.

    One box - thousands of servers.

    Huge savings.
     
    steve, Dec 18, 2006
    #12
  13. In message <>, Don Hills wrote:

    > In article <em5615$6i0$>,
    > Lawrence D'Oliveiro <_zealand> wrote:
    >
    >>These days, "capable" tends to mean "agile", that is to day "quickly
    >>adaptable to changing business requirements". That means heavier use of
    >>modern scripting languages and associated frameworks, which don't run very
    >>well on traditional mainframes.

    >
    > If it runs on Linux, it'll run on an (IBM) mainframe. ... Basically, with
    > a modern mainframe, you have a choice. You can either run many virtual
    > machines, one for each different application, or you can run one or two
    > big mutha apps - like running the application on a PC, but much, much
    > faster.


    But how much faster will it be, really? Remember that mainframes are
    optimized for bulk throughput, not quick interactive response. Whereas most
    things done with scripting languages are intended for interactive use, such
    as via the Web.
     
    Lawrence D'Oliveiro, Dec 18, 2006
    #13
  14. Lawrence D'Oliveiro

    Don Hills Guest

    In article <em5615$6i0$>,
    Lawrence D'Oliveiro <_zealand> wrote:
    >These days, "capable" tends to mean "agile", that is to day "quickly
    >adaptable to changing business requirements". That means heavier use of
    >modern scripting languages and associated frameworks, which don't run very
    >well on traditional mainframes.


    If it runs on Linux, it'll run on an (IBM) mainframe. Even if you're running
    Z/OS (was MVS), scripting is well taken care of. A large number of the
    queries in the comp.lang.rexx newsgroup are from mainframe programmers.
    Basically, with a modern mainframe, you have a choice. You can either run
    many virtual machines, one for each different application, or you can run
    one or two big mutha apps - like running the application on a PC, but much,
    much faster.

    --
    Don Hills (dmhills at attglobaldotnet) Wellington, New Zealand
    "New interface closely resembles Presentation Manager,
    preparing you for the wonders of OS/2!"
    -- Advertisement on the box for Microsoft Windows 2.11 for 286
     
    Don Hills, Dec 18, 2006
    #14
  15. Lawrence D'Oliveiro

    darkknight Guest

    On 18 Dec 2006 19:42:22 +1300, Dave Taylor
    <> wrote:

    >Lawrence D'Oliveiro <_zealand> wrote in
    >news:em55t5$6i0$:
    >
    >> About 450,000 servers
    >> <http://cbcg.net/talks/googleinternals/index.html>. The servers are
    >> cheap x86 boxes, except in one regard--they don't skimp on the power
    >> supply. That's because a good-quality supply is more efficient, which
    >> reduces power bills.
    >>

    >
    >For more on the GFS, (Google File System)
    >go here, and just search for Google, lots of interesting in depth talks on
    >how it really works from people who design and work with it:
    >Check out the CSE:
    >http://www.researchchannel.org/search/sitesearch.aspx
    >http://www.researchchannel.org/prog/displayevent.aspx?rID=4188&fID=345
    >
    >There at least two others in the CSE that have interesting info about how
    >Google runs its clusters.



    Well none of those links are any use. This one is much more
    interesting!
    http://www.google.com/technology/pigeonrank.html
     
    darkknight, Dec 18, 2006
    #15
  16. In message <>, Don Hills wrote:

    > In article <em5kmd$28t$>,
    > Lawrence D'Oliveiro <_zealand> wrote:
    >>
    >>But how much faster will it be, really? Remember that mainframes are
    >>optimized for bulk throughput, not quick interactive response. Whereas
    >>most things done with scripting languages are intended for interactive
    >>use, such as via the Web.

    >
    > A lot faster, really.


    I doubt that. In terms of sheer CPU power, there is no big gap between
    mainframes and PCs any more. The mainframes only have the edge in terms of
    I/O bandwidth.

    > A cluster of PCs for a given job will cost less than a mainframe,
    > but for a large cluster you need your own maintenance staff because
    > there's always something to repair.


    That's a reliability issue, not performance.

    > Say you have an intermittent
    > single-bit error in a memory stick in one machine of a 1000-node cluster.
    > It doesn't crash the machine, just gives altered data sometimes. If that
    > happens to give the wrong result in a critical query, you're screwed. You
    > have failure redundancy with a cluster, but no proper error checking. You
    > can specify ECC memory and full parity checking on all buses along with
    > redundant power supplies, but that adds an order of magnitude to the
    > hardware cost.


    In that case, it's cheaper to simply do the same calculation on at least 2
    machines. If they agree, accept the result. Otherwise, do it again. Cheaper
    than buying a mainframe.

    > Some of the higher price of a mainframe is in the spec of the components,
    > and a lot is in proper error checking and true redundancy.


    Again, that's a reliability issue, not one of performance.

    > When something
    > does fail it is almost always a partial failure - the failing CPU / bus /
    > drive/ memory card is placed offline and a service call placed by the
    > service processor unit, often without the operators noticing that
    > something has failed, and without losing the transaction that was in
    > progress at the time.


    There are transactional systems that achieve the same thing with massively
    parallel PC hardware. Such as what Google use, for instance.

    > On the software side, a single instance of an application, for example a
    > web application that makes heavy use of databases, is much easier to
    > administer, update and debug than a distributed version of the
    > application. Again, you pay more for the mainframe but less for the
    > programmers and analysts.


    On the PC side, there are frameworks that can take care of a lot of the
    details for you. This kind of thing has been researched for some decades
    now.
     
    Lawrence D'Oliveiro, Dec 18, 2006
    #16
  17. Lawrence D'Oliveiro

    Don Hills Guest

    In article <em5kmd$28t$>,
    Lawrence D'Oliveiro <_zealand> wrote:
    >
    >But how much faster will it be, really? Remember that mainframes are
    >optimized for bulk throughput, not quick interactive response. Whereas most
    >things done with scripting languages are intended for interactive use, such
    >as via the Web.


    A lot faster, really. Your view of a mainframe is somewhat dated. About 20
    years dated... even back in the 80s it was clear that the future was
    transactional processing rather than batch processing.

    A cluster of PCs for a given job will cost less than a mainframe,
    but for a large cluster you need your own maintenance staff because
    there's always something to repair. Say you have an intermittent single-bit
    error in a memory stick in one machine of a 1000-node cluster. It doesn't
    crash the machine, just gives altered data sometimes. If that happens to
    give the wrong result in a critical query, you're screwed. You have failure
    redundancy with a cluster, but no proper error checking. You can specify ECC
    memory and full parity checking on all buses along with redundant power
    supplies, but that adds an order of magnitude to the hardware cost.

    Some of the higher price of a mainframe is in the spec of the components,
    and a lot is in proper error checking and true redundancy. When something
    does fail it is almost always a partial failure - the failing CPU / bus /
    drive/ memory card is placed offline and a service call placed by the
    service processor unit, often without the operators noticing that something
    has failed, and without losing the transaction that was in progress at the
    time.

    On the software side, a single instance of an application, for example a
    web application that makes heavy use of databases, is much easier to
    administer, update and debug than a distributed version of the application.
    Again, you pay more for the mainframe but less for the programmers and
    analysts.

    --
    Don Hills (dmhills at attglobaldotnet) Wellington, New Zealand
    "New interface closely resembles Presentation Manager,
    preparing you for the wonders of OS/2!"
    -- Advertisement on the box for Microsoft Windows 2.11 for 286
     
    Don Hills, Dec 18, 2006
    #17
  18. Lawrence D'Oliveiro

    whome Guest


    >> A lot faster, really.

    >
    > I doubt that. In terms of sheer CPU power, there is no big gap between
    > mainframes and PCs any more. The mainframes only have the edge in terms of
    > I/O bandwidth.
    >


    Maybe PC's compete on massive parallel processing based tasks that require
    little disk IO.

    But, for a business applications mainframes win hands down.

    Horses for courses i guess.
     
    whome, Dec 18, 2006
    #18
  19. In message <>, whome wrote:

    >>> A lot faster, really.

    >>
    >> I doubt that. In terms of sheer CPU power, there is no big gap between
    >> mainframes and PCs any more. The mainframes only have the edge in terms
    >> of I/O bandwidth.

    >
    > Maybe PC's compete on massive parallel processing based tasks that require
    > little disk IO.
    >
    > But, for a business applications mainframes win hands down.


    Only certain kinds of business applications--those which are heavily
    transaction-based. Your bank is likely to need such a system, but you
    probably do not.
     
    Lawrence D'Oliveiro, Dec 19, 2006
    #19
  20. Lawrence D'Oliveiro

    whome Guest

    "Lawrence D'Oliveiro" <_zealand> wrote in message
    news:em7efv$bkk$...
    > In message <>, whome wrote:
    >
    >>>> A lot faster, really.
    >>>
    >>> I doubt that. In terms of sheer CPU power, there is no big gap between
    >>> mainframes and PCs any more. The mainframes only have the edge in terms
    >>> of I/O bandwidth.

    >>
    >> Maybe PC's compete on massive parallel processing based tasks that
    >> require
    >> little disk IO.
    >>
    >> But, for a business applications mainframes win hands down.

    >
    > Only certain kinds of business applications--those which are heavily
    > transaction-based. Your bank is likely to need such a system, but you
    > probably do not.


    agreed.
     
    whome, Dec 19, 2006
    #20
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. brian

    World's Largest Zoom Ratio

    brian, Apr 21, 2004, in forum: Digital Photography
    Replies:
    4
    Views:
    1,478
    Bob Hickey
    Apr 21, 2004
  2. Daniel Silevitch

    World's largest camera

    Daniel Silevitch, Jun 15, 2006, in forum: Digital Photography
    Replies:
    19
    Views:
    686
    George Kerby
    Jun 17, 2006
  3. Au79

    World's largest Wi-Fi network uses Linux

    Au79, Mar 15, 2006, in forum: Computer Support
    Replies:
    0
    Views:
    358
  4. Somebody

    Worlds Largest Photo and Worlds Largest Camera...

    Somebody, Aug 9, 2007, in forum: Digital Photography
    Replies:
    1
    Views:
    667
    John Turco
    Aug 16, 2007
  5. manazar
    Replies:
    0
    Views:
    1,510
    manazar
    Jan 31, 2008
Loading...

Share This Page