Forgot your password?
typodupeerror
Businesses Linux Business The Almighty Buck United Kingdom Upgrades Linux

LSE Breaks World Record In Trade Speed With Linux 452

Posted by timothy
from the linus-mentioned-their-startling-speed dept.
LingNoi writes with this excerpt from ComputerWorld UK: "The London Stock Exchange has said its new Linux-based system is delivering world record networking speed, with 126 microsecond trading times. The news comes ahead a major Linux-based switchover in twelve days, during which the open source system will replace Microsoft .Net technology on the group's main stock exchange. The LSE had long been criticised on speed and reliability, grappling with trading speeds of several hundred microseconds. The 126 microsecond speed is 'twice as fast' as its main international competitors, the London Stock Exchange said. BATS Europe and Chi-X, two dedicated electronic rivals to the LSE, are reported to have an average latency of 250 and 175 microseconds respectively. Neither company immediately provided details. But many of the LSE's older and more traditional rivals offer speeds of around 300 to 400 microseconds. Nevertheless, Linux is now standard in many exchanges, including the New York Stock Exchange."
This discussion has been archived. No new comments can be posted.

LSE Breaks World Record In Trade Speed With Linux

Comments Filter:
  • by Anonymous Coward on Monday October 25, 2010 @01:18AM (#34009198)

    Windows games: Plentiful and well developed. Mac games: Barely existent. Linux games: Lowest latency.

  • by QuoteMstr (55051) <dan.colascione@gmail.com> on Monday October 25, 2010 @01:18AM (#34009200)

    Trades that happen this fast are only good for further enriching large investment firms that can afford to spend millions on clever algorithms for shuffling numbers around. This speedup lets these companies make even more money without creating one damned thing that's useful to any living person.

    Limit trades to one per second per institution, and while you're at it, add that tiny per-trade tax. Finance should be boring. Let's encourage people to focus on the real economy that operates in the world inhabited by you and me.

    • by buchner.johannes (1139593) on Monday October 25, 2010 @01:23AM (#34009234) Homepage Journal

      Can someone explain to me what the system does in these 126 microsecond? Is it sending packets through the world, doing some complicated calculations, solving locking resources? It seems an awefully long time to add to a table and update some stats.

      • by Ethanol-fueled (1125189) * on Monday October 25, 2010 @01:28AM (#34009252) Homepage Journal
        The idiots in charge of the stock markets believe that, since Linux doesn't crash, their stock markets won't either.
      • by Anonymous Coward on Monday October 25, 2010 @01:40AM (#34009302)

        It is the time measured from when a bid/ask order is sent from the customer's network port, until it has been processed/stored and possibly matched at the Exchange, and back again.

      • Lots of data going in lots of data going out. Millions of people make trades on this thing constantly.

      • Re: (Score:3, Interesting)

        Anonymous Coward writes:

        It is the time measured from when a bid/ask order is sent from the customer's network port, until it has been processed/stored and possibly matched at the Exchange, and back again.

        ikkonoishi writes:

        Lots of data going in lots of data going out. Millions of people make trades on this thing constantly.

        nacturation writes:

        One eighth of a millisecond is an awfully long time?

        davester666 writes:

        It increases the time delta between when your broker trades the stock up a little, then puts through your trade by selling you the stock he just bought, but for a little more than what he bought it for. Every time.

        lena_10326 writes:

        You do realize this is a stock exchange processing large volumes of transactions requiring a high degree of availability and consistency? Not a ma-and-pop website processing 100 transactions a day. Right?

        Johannes concludes: no one knows what it does. Yet we are stunned of the millions of transactions flowing around.

        The limiting factor can't be throughput/number of transactions, given that the data flow is continuous.
        The closest to an answer is that it is network latency. But why is it relevant to use Linux then? Because the TCP/IP stack is so great? Wouldn't the specs of the network (fibre?) be more interesting?

      • by Anonymous Coward on Monday October 25, 2010 @05:01AM (#34010026)

        It's not only updating tables...

        The exchange gets a cut of every sale, so at the very least, the more sales you have the more you can profit.

        It's more insiduous than that though. Between the price a buyer wants to pay and a seller wants to receive, there's a certain spread. Faster trades allow the exchange to take advantage of that spread. Sometimes it lasts a second or so, but trading volume means a second can make thousands of dollars.

        Faster exchanges also allow 'tasting'. The average investor doesn't get to take advantage of it, but the large houses do. They can float a price out there and see how many people are willing to buy at that price. Then they can test a higher price... Then higher.. At some point they reach a price that maximizes their profit.

        It also means that certain brokerage houses can get their trades in faster. So that means a popular and rising stock goes to those houses that pay for the privilege of first dibs. These houses can then set the price on the stock.

        There are dozens of other ways that faster trades help.... Of course, none of it helps us, the average stockbroker.

      • 126 microseconds is the time light can travel about 38 kilometers. So NO, packets are not traveling far in that time. it's near the latency of standard Gig-E, so traders use Infiniband. And get closer to the exchange, as every 30 meters is a microsecond of advantage.
    • Re: (Score:3, Interesting)

      by mpoulton (689851)
      Trading this fast brings the market closer to optimal economic efficiency, where prices at any instant accurately reflect value. Latency contributes to the very inefficiencies that you blame these "large investment firms" from profiting off of. These high-speed arbitrage and "quant" investors may make a profit without creating a product (though collectively their track record is not very good financially), but their profit margins are vanishingly small and they serve a critical role in equalizing prices b
      • by dakameleon (1126377) on Monday October 25, 2010 @01:50AM (#34009334)

        I respect your optimism and idealism when it comes to these things, but Index Arb desks are making some of the most effective, near-risk-free profits for the big banks, and it's little wonder the LSE wants to be at the forefront of this market. You don't have to pay a computer a bonus, and the programmers behind this hardly see the same kind of money as the big swinging dick traders who try to spot the macro inefficiencies. Furthermore, the same strategies and speed advantages are used for algo traders to allow big blocks of trades to go through as best possible without shifting the market, making more cash when the trades are billed to the client at a weighted average instead of the true cost.

        But then you don't see these numbers in the breakdown of the Goldmans profit numbers, and you never will. In the casino of the share market, the dealer is helping the sharks fleece the sheep.

      • by NoSig (1919688) on Monday October 25, 2010 @01:55AM (#34009352)
        I can't tell if you are are someone defending your own work to assuage your conscience or a troll pretending to do so, but anyway: There is no value to equalizing prices in less than a second, especially when the "equalizing" you are talking about is really just pocketing the difference. The only value there would be in equalizing prices in less than a second would precisely be to remove the threat that these people pose. Since sub-second trades are necessarily automated trades, they also cannot be doing anything sensible to keep prices up to date as real world conditions change, as that requires understanding the real world which an automated system cannot. They serve only the function of increasing profits for their investors - how little or how much damage they cause in the course of that is what is debatable.
      • by atomic brainslide (87546) on Monday October 25, 2010 @02:22AM (#34009454) Homepage

        while in theory your idea is correct, the harsh reality is that in practice, the large investment firms increase their profits drastically because there are actually two markets. this isn't strictly legal, but it's there. the large firms have dedicated connections to the exchanges with guaranteed SLAs and lower latencies than any other regular participant in the market. this allows them to stuff the buy/sell queues and rapidly cancel orders before they go through. the purpose of this is to deduce other bidders' price points and gain an edge. there are a number of such hedge funds (and even a major bank whose name escapes me), for example, that have had perfect trading days for over a year. statistically impossible outcomes like this only come from gaming the system in the above mentioned manner. as usual, the regulators are asleep at the wheel and the markets become more volatile week to week with increasing flash-crashes exactly because of these schemes. more efficient markets these are not.

      • Re: (Score:3, Informative)

        by nacturation (646836) *

        Trading this fast brings the market closer to optimal economic efficiency, where prices at any instant accurately reflect value. Latency contributes to the very inefficiencies that you blame these "large investment firms" from profiting off of.

        The faster the trades, the greater the disparity between how fast a normal Joe investor can trade vs. how fast an investment firm plugged directly into the exchange can trade. Let's say I'm down the block and my ping times are an awesome 6 milliseconds. A large investment firm plugged into the exchange might have a ping time of 1 millisecond (I'm being quite conservative here), even 1% increase in benefit I see the investment firm is going to see a 6% benefit.

        Yes, they play a role in equalizing market pri

      • by r00t (33219) on Monday October 25, 2010 @02:43AM (#34009528) Journal

        Prices are discrete, quantized, non-differentiable, etc.

        Prices are chaotic and somewhat fractal.

        Going faster does not solve this. Think of sign(sin(1/x)) as x approaches zero; it changes rapidly but this doesn't make it smooth.

        Hourly trades would be reasonable. You get a few minutes to submit secret bids, the exchange gets nearly a half hour to match them up, the exchange gets a few minutes to publish results, and you get nearly a half hour to decide on your next bid.

        There is no reason that the finances of normal corporations and normal investors should be subjected to the abuse of today's stock market.

      • by Tom (822) on Monday October 25, 2010 @02:54AM (#34009566) Homepage Journal

        Trading this fast brings the market closer to optimal economic efficiency, where prices at any instant accurately reflect value.

        It might, if the trading were actually being done at the exchange, for market prices. It isn't. Most trading today is only registered at the exchange, but the actual deal is brokered elsewhere.

        Also, the amount of liquidity a market needs is subject to discussion. Do you really need to have a counterpart available this second for a market to work? That is nonsense, the market (the real market, not the speculative one) wouldn't burn and die if you had to wait a second or two or even *gasp* five for a deal to go through.

        Providing liquidity is a valuable role. However, you are ignoring the fact that everyone is in it for the money. If the cost of liquidity, i.e. the amount of profit the high-speed traders extract from the market, becomes too high, the market also suffers. Somewhere, there is an optimal point between the positive (more liquidity) and the negative (the cost of this liquidity).

        • by Lobachevsky (465666) on Monday October 25, 2010 @09:22AM (#34011748)

          Liquidity isn't just about there being _someone_ willing to buy or sell; it's about the spread. Do you want to go back to 25 cent spreads from the early '90s? Most spreads today are 1 cent. If you're happy paying 25 cent spreads to get rid of automated traders, I'd say that's a bit like chewing off your arm to swat a fly. Most automated traders make anyways from 0.1 cents to 0.5 cents per share. Comparatively, retailers like E*Trade charge customers $9.99 for trades that average around 400 shares, which is 2.5 cents per share. Mutual funds like Fidelity often charge 1 to 2% management fees on investment, which is 30 to 60 cents per share on a $30 stock.

          Trying to bend the rules of the market to wipe out a segment that makes 0.1 to 0.5 cents per share is silly when there's zero effort concerning E*Trade making 2.5 cents per share, or Fidelity making 30 to 60 cents per share. Professional services like Lime (a high-end version of E*Trade) charge 0.1 cents per share. No one is angry that E*Trade charges 2.5 cents per share while Lime charges 0.1 cents per share? Oh, that's right, because it's "only $9.99 !!"

          That's the irony of all of this. The average person writes of $9.99 to E*Trade but the media tries to get them concerned about "costs" that are effectively 1/20th of that. I put quotes around "costs" because the spreads have come down from 25 cents per share in the early '90s to 1 cent nowadays. So the average person benefited 24 cents on the spread, and is angry that liquidity providers make 0.1 cents per share? Where was the anger in the early '90s when specialists (a cartel of liquidity providers) were raking it in, making 10+ cents per share? Where is the anger now at Fidelity _losing_ our money in 401ks _and_ charging management fees of 30 to 60 cents per share, annually?

          The biggest to benefit from automated traders going away are the Larry Ellisons and other market manipulators who want to buy up companies without "moving the price up." It's all misdirection, trying to convince you and me that we're being hurt. It's the "death tax" all over again. The average person was never affected by estate tax, yet the media convinced us we should be against it, just so some rich jackasses can save money. The same deal is happening now. Rich folks want to buy out the public companies you and I are invested in, cheaply, and stealthily. They don't like automated traders sniffing their actions with pattern matching heuristics and raising the prices (benefiting long-term owners like us).

      • by hughbar (579555) on Monday October 25, 2010 @02:59AM (#34009586) Homepage
        There's a huge problem with the word 'value' in the above.

        To declare interest, I'm ex-investment banking and not too proud of it. The 'values' and 'derivatives' exchanged are often not mapped to anything happening in the 'real' world [think manufacturing, [useful] services even], however they do have a negative impact on it [factory closures, bank bailouts paid by the taxpayer, for example].

        Try a thought experiment, does anything useful change in 126 microseconds? Bread get baked? Pizza cooked? House built? Seed planted, if you want to get rural and idyllic?

        Incidentally, I'm not against simple futures, for example, they smooth the farmer's year and have a purpose. I am pretty much against most exotic financial derivatives and against short-term [126 microseconds, for example!] 'investment' to use ironic quotes...
        • Re: (Score:3, Insightful)

          by gundersd (787946)

          It may be true that nothing useful in the real-world changes in 126 microseconds... however...

          The way I look at it is that 'the market' is like a big complicated electronic system which contains a lot of complex feedback loops (some of them more stable than others). Imagine tweaking a random knob on such an electronic circuit and watching the effects of that tweak ripple through the circuit until it (hopefully) reaches a steady state again.

          Increasing the latency causes changes to ripple through the syst

        • Re: (Score:3, Insightful)

          by Rockoon (1252108)

          Try a thought experiment, does anything useful change in 126 microseconds? Bread get baked? Pizza cooked? House built? Seed planted, if you want to get rural and idyllic?

          But isnt that the point?

          I would think that an exchange where useful things CANT happen between the decision to buy and the actual fulfillment of the buy would be a desirable one.

        • by khallow (566160) on Monday October 25, 2010 @09:25AM (#34011810)

          Try a thought experiment, does anything useful change in 126 microseconds? Bread get baked? Pizza cooked? House built? Seed planted, if you want to get rural and idyllic?

          Yes. All of the things you mention, for starters. I'm being a bit pedantic, but you do any of those activities in a long string of periods of 126 microseconds.

          It surprises me how many people do a job for years and fail to understand what it is they do. You should know what a baker, pizza maker, home builder, and farmer have in common. They all trade. People like derivatives traders take that to the purest level, trading fairly abstract things, but still things with an attachment, however tenuous, to the real world. Nothing wrong with that. The trades themselves don't close factories or cause bank bailouts.

          The real problem is leverage. For example, when you were a banker, how much collateral did you have to put up for your borrowed money? 1 unit per 10 of debt? More? In the real estate securities market, there were apparently people who could borrow 50 units for every unit of assets they had. I can't comprehend that level of foolishness, though I'm sure it make good bank for a time for the people who could get away with it.

          Bottom line, what do you think would happen if pizza makers could get 50 to 1 leverage? I could use the pizza store I owned to borrow enough to build 50 more. I could use my car as collateral for one or more pizza stores (depending how nice it is). In a few months, we'd be up to our eyeballs in new pizza restaurant construction. There'd be incredibly specialized stores catering to the gay, vegan, Hispanic boardgamer.

          And after the bust, when people realize that they didn't like pizza all that much? Pizza would be vilified, an epithet for people who need something to hate. Pizza makers would be scoundrels of the Earth who don't make anything connected to the real world, unlike bread bakers, home builders, farmers, or even the investment banker. Much would be made of their negative impact, such as factory closures and bank bailouts (paid by the taxpayer), for example. Even pizza makers would be self-flagellating themselves over their worthlessness.

          The above is an interesting sociological phenomena, but the lesson boils down to high leverage causes massive fuckups every time no matter the industry, no matter how "real" the product is, and is the number one cause of market crashes and trigger for recessions.

      • Planet Money (Score:3, Interesting)

        by VValdo (10446)

        Trading this fast brings the market closer to optimal economic efficiency, where prices at any instant accurately reflect value. Latency contributes to the very inefficiencies that you blame these "large investment firms" from profiting off of.

        Let's just relax and listen to this episode of Planet Money on high-frequency trading [npr.org].

        I know. It's NPR. They fired Juan Williams. Whatever. (FWIW: on the Diane Rehm show from Friday, they defend Williams, say there was a "rush to judgement" and say that NPR went " [youtube.com]

      • by Pinky's Brain (1158667) on Monday October 25, 2010 @03:25AM (#34009684)

        The HFT algorithms aren't value based, how could they cause the stocks to accurately reflect value? HFT is a pure arbitrage play, gaming the restricted strategies of long term investors (in some cases legally restricted). Forcing volatility beyond normal variance to try to benefit from stop loss orders is a perfectly valid HFT strategy ... does nothing to make stocks reflect value though.

        HFT has one benefit, increased liquidity, and one downfall, increased volatility. For the moment humans are still the only ones capable of judging value ... the HFT algorithms add nothing there.

      • Re: (Score:3, Interesting)

        Trading this fast brings the market closer to optimal economic efficiency, where prices at any instant accurately reflect value.

        This is fundamentally wrong. The only way to bring the market mathematically closer to the ideal is to speed up time for everybody in the market. But that's certainly not what's going on.

        Only some market participants get to trade at a faster rate in reality, so what this leads to is a bias in the prices which weighs the preferences of those privileged participants more than th

    • by 1s44c (552956) on Monday October 25, 2010 @02:14AM (#34009412)

      Much as I'm sure that Linux will perform faster and with far more stability than Windows in all given applications the parent is spot on here.

      More speed isn't the answer here, less speed is needed. If anyone wants to buy and sell little bits of companies at those kinds of speeds they clearly have no interest whatsoever in the companies they are trading. They can only be gaming the system for unearned gain.

      Investment bankers ( and I don't mean real investors ) are parasites on the capitalist system and anything that enriches them impoverishes someone who works for a living.

    • THIS (Score:5, Insightful)

      by Joce640k (829181) on Monday October 25, 2010 @04:08AM (#34009804) Homepage

      If you want to stabilize the global economy put a tax on all stock trades. Stocks and shares should be long term planning, not microsecond.

      • Re:THIS (Score:4, Interesting)

        by asc99c (938635) on Monday October 25, 2010 @06:46AM (#34010482) Homepage

        Yet oddly, in the UK at least, there is a tax on actually buying shares, so my long-term investment account has various charges for stamp duty as my money goes in there. The tax is only applied to buying actual shares. It isn't applied when messing around with futures and complex derivatives etc that have caused the problems.

        I think this needs to be exactly the opposite way around. I don't think there should be a tax for dealing in shares, but instead a tax on dealing with various futures products. Not too severe a tax but enough to dissuade this sort of high speed speculation.

    • This stupid system makes it possible to earn a massive amount of money by buying and selling shares at the right time. The really stupid thing is that at some point, the money earned with those transactions can be exchanged for real goods, produced by real people.

      People who don't add anything tangible to this world get free money, and can buy goods for it. It's madness. If I hadn't grown up in this world, I doubt that I would ever understand and accept it...

    • Re: (Score:3, Insightful)

      by rtb61 (674572)

      So the London stock exchange has updated it's operating system to accelerate it trades, just as it seems likely that laws will be implemented to decelerate trades. It makes logical sense that trades be slowed to a 24 hour cycle to stabilise the global share market a prevent algorithmic stock price collapse or major firms distorting the market electronically with rapid internal trades.

    • Re: (Score:3, Insightful)

      by hoggoth (414195)

      You cannot come up with a procedure or technology that will defeat greed.
      One trade per second per institution? They will register 1,000 institutions and have their trades auto-distributed throughout their sub-companies.

  • Software Only? (Score:3, Insightful)

    by ticklish2day (575989) on Monday October 25, 2010 @01:20AM (#34009212)
    Is this improvement purely because of the change in software technology or were there simultaneous infrastructure and process modifications? The article doesn't really say.
  • Linux: 1, MS: -1 (Score:3, Interesting)

    by ThePromenader (878501) on Monday October 25, 2010 @01:23AM (#34009232) Homepage Journal

    Not so sure that the networking speed is due only to the Linux OS switchover - The LSE obviously updated its hardware too.

    I'd be really be interested in the makeup of the LSE network support - do they rely on their own developers/deployers, or do they have a support deal with Linux?

    • by LingNoi (1066278)

      I found this further into some other links about it.

      At the time the LSE maintained that TradElect was not responsible for the outage, but has since, nevertheless, made aggressive steps to replace the platform by acquiring trading firm MillenniumIT, the supplier of its new system.

      I'd post more info but slashdot has broken copy and paste.

    • by Anonymous Coward on Monday October 25, 2010 @01:57AM (#34009362)

      It was costing them so much to maintain their systems due to support and modification contracts that they just out and out bought an ENTIRE company whose sole product was...trading systems (For about 50 million'ish pounds IIRC).

      In essence they bought a development department lock stock and barrel and it was STILL cheaper than their existing setup.

  • by ClarkMills (515300) on Monday October 25, 2010 @01:24AM (#34009236)

    http://www.youtube.com/watch?v=BwSM55bsCrM [youtube.com]

    I could watch it over & over... It puts a smile on my face... :)

    http://news.cnet.com/8301-13846_3-10036286-62.html [cnet.com]

    Cheers... Clark

  • by Anonymous Coward on Monday October 25, 2010 @01:25AM (#34009238)

    Communist Linux - bringing speed and efficiency to the dregs of capitalism.

    • by shutdown -p now (807394) on Monday October 25, 2010 @03:30AM (#34009706) Journal

      It all goes as planned, comrade. Remember: the faster the capitalists trade, the quicker they will sell us the rope with which we will hang them! ~

  • Gee (Score:5, Insightful)

    by TubeSteak (669689) on Monday October 25, 2010 @01:26AM (#34009242) Journal

    The record breaking times were measured on the LSE's Turquoise smaller dark pool trading venue, where trades are conducted anonymously.

    Dark pools are part of the problem.
    Transparency is critical for a functional marketplace.
    Dark pools only require trades to be listed after the fact...
    Which isn't as useful as it sounds, even if brokerages weren't completing the trades in/across-house, where disclosure is not required.

    Anonymity and secrecy are anathema to a functional market

  • It may be that a Linux network is able to conduct trades faster than a windows one. I'd be willing to be though that most of this dramatic speed increase is down to new hardware. The windows network the LSE was on was what... 4+ years old?
  • Hip hip hooray for Wine!

    I'm sure they took the complete system as it used to run on Windows and then simply ran it on Linux using Wine, thus allowing for great comparison.

    Or maybe they completely redesigned the system, avoided known bottle necks and halved the response time.

    I'm a UNIX developer since many, many moons but I find these stories lacking detail highly disturbing. Sure, nice Linux PR but on /. I'd hope to get a bit more context information.
    • Seriously? You seriously think they would use Wine for a real time system? Are you out of your mind? Much more likely what they did is they ported over a lot of their code to run on a heavily modified real-time Linux(whose scheduler is a hell of a lot different than the one used by desktop distros). If you know what you are doing writing portable C isn't all that hard.
  • trading speeds of several hundred microseconds

    Several hundred? All of the stories I've seen in the past about TradElect indicated that trades were measured in the milliseconds... The new system seems to be about twenty times faster.

  • by mathfeel (937008) on Monday October 25, 2010 @01:30AM (#34009268)
    1. Create a super fast OS that take over the world's stock trading
    2. Ruin the economy so that no one can afford to buy proprietary product any more
    3. ...
    4. Linux on every desktop!!!
  • by pgn674 (995941)

    Distance light travels in 126 microseconds: 23 miles, or 38 kilometers. That's fast.

    TFA isn't detailed on what exactly takes 126 microseconds, calling it 'trading time,' 'trading speed,' 'latency,' and maybe how long it takes to process orders. I'm wondering, where are the computers that are doing the trading? Do they need to be positioned within a few miles of the stock exchange servers, with dedicated fiber optic lines between them? Because it sounds like that'd be necessary to make this kind of speed up

    • by Anpheus (908711)

      If it's anything like the NYSE is planning, they let the investment banks build their high speed trading server racks right next to the stock exchange racks. Literally.

      Supposedly there's even ways to get your server physically in the same rack as one of the NYSE's exchange servers, but I think that might be taking it a bit too far, or maybe an overzealous reporting on the matter. But I don't think it's even disputed anymore that they're letting banks colocate with the exchange.

      • Re:Light (Score:5, Insightful)

        by Nursie (632944) on Monday October 25, 2010 @02:16AM (#34009422)

        The world of high finance has become just a bunch of racks frantically swapping bits around.

        Yet when it screws up, the shocks are felt everywhere, for some reason.

        Something is sick on this planet, when automated behaviour of electronic systems decide who eats, who can buy a new mansion, who gets a miserably low pension, whose house is going to be taken away and who's going to pocket a billion dollars in profit.

        I'm more and more coming over to the side of those that say the whole finance sector is parasitic in nature and needs to be destroyed.

    • Re:Light (Score:5, Informative)

      by dakameleon (1126377) on Monday October 25, 2010 @01:59AM (#34009366)

      Yep, and major trading firms will do anything to get closer to the exchange servers. I worked for a Major Bank which had a major data centre well outside London which hosted all the "slow" apps, and a small (but well cooled) server room in the City a few blocks from the LSE building. Each and every app in the central data room had to justify its need, and every so often you would hear about acquisitions of real estate closer still.

      This is of course pre-2008; I'm no longer so intimate with the details of server rooms at major banks. C'est la vie.

      • Re:Light (Score:4, Interesting)

        by dintech (998802) on Monday October 25, 2010 @06:34AM (#34010446)

        Co-location. A lot of exchanges will let you put your servers in their building. Then the arguments start about who's network cable is shortest and which router it's on. I'm not joking.

        • Re: (Score:3, Insightful)

          by TubeSteak (669689)

          Co-location. A lot of exchanges will let you put your servers in their building. Then the arguments start about who's network cable is shortest and which router it's on. I'm not joking.

          Ultimately, the solution to all this is to give the exchange a heartbeat and to only execute trades on those beats.
          Randomize the queued orders that accumulate between the beats and we can finally kill off high frequency trading.

          Equal access is one of the prerequisites for a perfect market [wikipedia.org]

  • trade speed (Score:5, Insightful)

    by Eivind (15695) <eivindorama@gmail.com> on Monday October 25, 2010 @01:41AM (#34009304) Homepage

    Trade-speed is irrelevant to investors.

    It's relevant to highly speculative robot-trader algorithms that try to make a profit by arbitraging sub-second timing-issues. But this is a zero-sum game: one trader can only gain $X by taking advantage of timing if other traders lose PRECISELY $X, so to the sum of traders, this is irrelevant.

    Stock-exchanges, make a living trough fees. The fees are coupled with volume, i.e. a broker that has a larger volume of orders, will pay higher fees.

    So lower latency is good for the stock-exchange, neutral for traders on equal grounds and negative for those suckers who play at daytrading. It -does- tilt the table towards those with machinery though, but the effect is irrelevant for traders who aren't extremely short-term.

    In short: yet another reason to invest rather than speculate.

    If you buy and sell 20 times today, each time with the table tilted a tenth of a promille against you, you'll on the average lose 2 promille, plus the fees. This doesn't sound like much, but a trader that does this 200 days a year, will have lost 20% of his profit to the tilted table. (if his flat-table profits where less than 20%, he'd thus run a minus)

    Meanwhile, the investor, who holds stock on the average 5 years, will also lose a tenth of a promille in every transaction, but since he's got 2 transactions in 5 years, that works out to 0.4 transactions/year -- thus his loss relative to the flat table is 0.4 * 0.01% = 0.004% pro year, which is irrelevant.

  • Does it run LINUX? Oh wait, it does. Cool!

  • by Idaho (12907) on Monday October 25, 2010 @02:01AM (#34009372)

    Remember that this very stock exchange moving to a purely Microsoft/.NET based solution was widely touted in Microsoft's so-called 'Get the "Facts"' campaign. Microsoft was involved (with Accenture) in the implementation of the project, not just in selling some Windows licenses. So this screwup should really be a PR disaster for them. If Microsoft themselves cannot even get a .NET project to work in places where their Linux-using competitors have no trouble at all (Chi-X is also Linux-based), then that sure looks like a platform in trouble to me.

    Remember that the entire thing crashed down for an entire trading day, something that you can imagine didn't go over well, and together with the high latencies and other numerous problems, was the reason they dumped it for Linux.

    • Re: (Score:3, Insightful)

      by wrook (134116)

      As much as I want to agree with you, it's really pure speculation. Projects fail for a huge number of reasons of which computing platform is just one. It might be a PR disaster, but not for any legitimate reason. I daresay it is not impossible to build a decent trading system on Windows/.NET. I have implemented systems in .NET before and while it's not my platform of choice I don't really recall anything about it that should be a show stopper here.

      But while we're speculating, it would be interesting to

  • by bertok (226922) on Monday October 25, 2010 @02:05AM (#34009382)

    I'm willing to bet that this has little to do with Linux.

    It's a great server OS, sure, but lets look at this realistically:

    - the Windows / .NET trading system was based on Windows 2003 and SQL 2000, and was deployed in 2005.
    - the Linux-based system is under development now, to be deployed next year.

    Just based on that, you'd expect substantial performance differences from just using newer hardware. Chances are that the original kit was certified as a part of the solution, and hasn't been replaced since. With all-new gear, I'd expect about 2-3x the CPU performance, and if they're using 10GbE (likely), then 10x the network performance.

    Even ignoring the hardware and the OS, one would expect 90% of the performance to be determined by the application, not the OS. Decisions like writing the software in .NET versus C or Java, or using a special-purpose Java runtime would make a huge difference, irrespective of the OS.

    On top of this, the software stack is completely different, and developed by a different team. Just about every design decision, small and large, will be different.

    To make a fair comparison, you'd have to run the new software on both OS platforms, on the same hardware.

    • by schon (31600) on Monday October 25, 2010 @08:11AM (#34011000)

      It's a great server OS, sure, but lets look at this realistically:

      - the Windows / .NET trading system was based on Windows 2003 and SQL 2000, and was deployed in 2005.
      - the Linux-based system is under development now, to be deployed next year.

      You missed

      - the Windows / .NET trading hardware has been upgraded continuously because it was unable to cope with the load.

      Just based on that, you'd expect substantial performance differences from just using newer hardware.

      Sure, except for the part that the both are running on new hardware.

      Chances are that the original kit was certified as a part of the solution, and hasn't been replaced since.

      "Chances are" - except that is 100% wrong. They had problems since day 1, which were blamed on the hardware, so they've been constantly upgrading it trying to fix the problem.

      Even ignoring the hardware and the OS, one would expect 90% of the performance to be determined by the application, not the OS. Decisions like writing the software in .NET versus C or Java, or using a special-purpose Java runtime would make a huge difference, irrespective of the OS.

      The old system was written with the help of MS. They were the ones that said that .NET was the best way to implement it, and they even touted this in their press releases.

      On top of this, the software stack is completely different, and developed by a different team. Just about every design decision, small and large, will be different.

      Of course it's completely different - that's the entire fscking point.

  • Great news: now the economy can crash even faster !!

  • by Tom (822) on Monday October 25, 2010 @02:46AM (#34009538) Homepage Journal

    One shouldn't forget that these high speed tradings are not in the interest of investors, companies being traded or other market participants with a real-world interest. They serve only the high-speed/high-volume speculative traders, specifically algorithmic or automated traders.

    Even Wall Street is slowly waking up to the problem:
    http://online.barrons.com/article/SB50001424052970203952604575552190237324972.html?mod=BOL_twm_mw [barrons.com]

    This was a bomb about a week ago when it was published, the guy making those statements is a Wall Street billionaire, not a hippie communist.

  • Markets (Score:5, Funny)

    by captain_dope_pants (842414) on Monday October 25, 2010 @04:50AM (#34009982)
    The stock market used to be full of bears & bulls. Now it has penguins too :)

This file will self-destruct in five minutes.

Working...