Forgot your password?
typodupeerror
Unix Operating Systems Software Linux IT

NY Stock Exchange Moves To Linux 272

Posted by kdawson
from the bye-bye-big-iron dept.
An anonymous reader writes "Even the old mainframe strongholds, the financial markets, are moving away from big iron. The New York Stock Exchange is one of them, as it's leaving the mainframe for AIX and Linux. They're doing it to save money; it seems that transactions are going to cost half as much on Unix and Linux as they did on the mainframe." The first phase of the transition happened last Monday.
This discussion has been archived. No new comments can be posted.

NY Stock Exchange Moves To Linux

Comments Filter:
  • by Timesprout (579035) on Thursday May 17, 2007 @10:03AM (#19160975)
    The bulk of the savings seem to be coming from reduced hardware and maintenance costs by getting rid of the mainframe and the savings are the reason they are doing it.
  • hmm (Score:4, Interesting)

    by nomadic (141991) * <nomadicworld&gmail,com> on Thursday May 17, 2007 @10:06AM (#19161059) Homepage
    What about stability issues? I'd think that these machines would have to be a little bit more robust than linux is capable of being at the moment.
  • by brunascle (994197) on Thursday May 17, 2007 @10:08AM (#19161117)
    ctrl-f tells me they didnt mention microsoft or windows. if it wasnt on *nix before, what was it on?
  • Well, the migration strategy seems interesting, although not especially surprising; they've eschewed emulation strategies that might incur a performance penalty in favor of some company that actually recompiles the old COBOL and IBM JCL code for modern architectures and does a lot of in-house QA (and, one assumes, has really good support...). They're using smaller IBM AIX servers to actually run the code in the new system, with the HP Linux machines basically doing all the I/O and general feeding of data.

    I'm a little surprised that IBM didn't manage to sell them on a new mainframe, or at least on its own clustered solution; or that they didn't ditch IBM completely and go with somebody else (what I'd suspect if somehow someone at IBM had really stepped on the wrong foot).

    There's not a whole lot of information in TFA about their old system, which actually sounds like it must be fairly neat; it's only described as a "1,600 MIPS mainframe" and then from context it's clear that it's an IBM of some sort. Another surprising thing is that they complain that the software licenses for it, among other things, are prohibitively expensive -- you'd think that IBM, in danger of losing a mainframe customer completely to commodity kit, would cut them some sort of a cheap-or-free deal on the software just to keep them around and on the support contracts. (I really gotta wonder if someone really boned this up; I mean, if you can't keep a mainframe contract at a place like the NYSE, really, what are you doing?)
  • by smooth wombat (796938) on Thursday May 17, 2007 @10:22AM (#19161361) Homepage Journal
    To test Clerity, Feldman and his team collected some batch and CICS application code, sent it to Clerity and said he would be at Clerity's development center in 24 hours to see the results. Clerity passed.


    Now that's service. I realize it's only compiling one code into another form but being able to take the code, compile it into what you need AND still have it work correctly in a 24 hour period is no easy feat.

    If nothing else, other firms will look at this migration to an aix/linux platform and see the cost benefits of doing so. After all, if the NYSE has done it, it can't be a bad thing.

  • Based on other parts of the article I assumed that it was an IBM mainframe of some sort; at ~1200 MIPS it must have been a fairly big one, although one assumes nothing too recent, or they probably wouldn't be spending a whole lot of money to replace it right now.

    I'd be interested to know if someone has any information or more educated guesses on what they probably have.

    (Humm ... wonder how they'll get rid of it. Everybody keep your eye on eBay!)
  • Re:Licensing Fees (Score:5, Interesting)

    by Clover_Kicker (20761) <clover_kicker@yahoo.com> on Thursday May 17, 2007 @10:26AM (#19161443)
    IBM will support your old hardware almost forever.

    They don't enjoy it, though - they have to stock a zillion old parts for a zillion old architectures, they have to train new guys on stuff that was obsolete before they got out of diapers.

    They gradually crank up maintenance fees to "encourage" you to upgrade to new kit that is easier to support.

  • by bananaendian (928499) on Thursday May 17, 2007 @10:46AM (#19161783) Homepage Journal

    Enough of the ignorant "Linux is the greatest!" drivel...

    They made a bad financial desicion.

    Any savings they think they made in the hardware, licensing and support costs will be lost many times over as soon as the system makes a small error or goes down.

    This is the reason why financial transactions still use mainframes.

    Mainframes [wikipedia.org] are unique in their integration and optimization between the hardware and the operating system they run. It gives you a level of performance, integrity and fault tolerance which cannot be achieved by taking generic hardware and sticking Linux on top.

    They probably thought (ironically like the stockmarket does) that in the short term there is significant savings to be made switching the system to generic hardware and linux. In the long term however they will be faced with more expensive and frequent maintenance and upgrade cycles. Mainframes on the otherhand are scalable almost to infinity and you pay for the reliablity and maintenance upfront when you purchase the system.

    What they could've done is buy a new mainframe and run their application level with virtualized Linuxes. After all this is what mainframes are good at.

  • Too bad... (Score:3, Interesting)

    by wandazulu (265281) on Thursday May 17, 2007 @10:54AM (#19161947)
    I worked at a world-wide bank that ran its entire operations on a pair of ES/9000 mainframes from IBM and, while insanely expensive, requiring a full-time staff of 12 people (for each machine), requiring a separate floor on the building, etc. etc., I have still never seen anything that came close to the horsepower these things had. They simply wiped the floor with everything else out there.

    As an example, it calculated a person's balance by starting with their opening balance, then went down the vsam file, adding and subtracting amounts, till it reached the bottom and gave the total. This process was instantaneous, even given all the other things it was doing.

    Sure there are better ways to do it, like storing the data in a real RDMS, using a trigger to update a "balance" field so it's a quick query instead of a lot of calculations, etc., but I wonder if so much of what we do is simply making the best of essentially a hardware deficiency; the baddest Intel-based Linux box probably couldn't do what this 20 year old mainframe can do, so we make it do the same thing but in different ways.

    Working with the mainframe programmers, all Cobol folks, made me think always of that great Dilbert cartoon of the smug Unix guy giving Wally a nickel and saying "get yourself a real computer" ... these people did not worry about efficiency for the most part simply because the machine was so fast, they didn't need to.

    So ultimately it's too bad that mainframes, for all their horsepower, really do resemble, to a certain extent, the moniker "dinosaur" in that their mammoth bulk simply couldn't get them out of the tar pits of cost and space.

    The coda to this is that, once you've used JSO on TSO, every Unix command looks like it's written in the Queen's English by comparison.

While money can't buy happiness, it certainly lets you choose your own form of misery.

Working...