Forgot your password?
typodupeerror
Network Japan The Internet United Kingdom Technology

$1.5 Billion: the Cost of Cutting London-Tokyo Latency By 60ms 158

Posted by Soulskill
from the aussies-still-jealous dept.
MrSeb writes "Starting this summer, and thanks to the continuing withdrawal of Arctic sea ice, a convoy of ice breakers and specially-adapted polar ice-rated cable laying ships will begin to lay the first ever trans-Arctic Ocean submarine fiber optic cables. Two of these cables, called Artic Fibre and Arctic Link, will cross the Northwest Passage, which runs through the Canadian Arctic Archipelago. A third cable, the Russian Optical Trans-Arctic Submarine Cable System (ROTACS), will skirt the north coast of Scandinavia and Russia. All three cables will connect the United Kingdom to Japan, with a smattering of branches that will provide high-speed internet access to a handful of Arctic Circle communities. The completed cables are estimated to cost between $600 million and $1.5 billion each. As it stands, it takes roughly 230 milliseconds for a packet to go from London to Tokyo; the new cables will reduce this by 30% to 170ms. The latency drop will mainly benefit algorithmic stock market traders, but other areas like education, telemedicine, and POTS will also enjoy the speed-up. Perhaps more importantly, almost every cable that lands in Asia goes through a choke point in the Middle East or the Luzon Strait between the Philippine and South China seas. If a ship were to drag an anchor across the wrong patch of seabed, billions of people could wake up to find themselves either completely disconnected from the internet or surfing with dial-up-like speeds. The three new cables will all come down from the north of Japan, through the relatively-empty Bering Sea. In addition, the Arctic Ocean, where each of the cables will run for more than 5,000 miles, is one of the least-trafficked parts of the world. That said, the cables will still have to be laid hundreds of meters below the surface to avoid the tails of roving icebergs."
This discussion has been archived. No new comments can be posted.

$1.5 Billion: the Cost of Cutting London-Tokyo Latency By 60ms

Comments Filter:
  • Bering Sea? (Score:5, Funny)

    by wolrahnaes (632574) <sean@seanharlow[ ]fo ['.in' in gap]> on Wednesday March 21, 2012 @12:14AM (#39423213) Homepage Journal

    Next, on Discovery's Deadliest Cabling Job...

    • by unixisc (2429386) on Wednesday March 21, 2012 @12:38AM (#39423397)

      Why not just leave the latencies alone? It's not like one gets cumulative latencies - the 230ms is constant in time.

      I guess the Bering Sea will be the crosspoint b/w the Siberia-Alaska railline that the Russkies want to build, and this cable that runs from the Arctic to the South. Probably run it along the Kamchatka peninsula coastline, then across to Sakhalin, Japan, then on to Taiwan, Philippines and along the S China Sea to Singapore on one end, and on the other, from Philippines, run it along to Papua New Guinea and then Australia and New Zealand. From Singapore, they could run a line to India, and get enhanced bandwith in that country.

      • You get the added benefit of not routing your economy through the NSA.
      • by necro81 (917438)
        The length of the cable is significantly shorter (by thousands of km), therefore the latency is pretty much guaranteed to be shorter. I don't think latency is as big a motivation as some would believe, but reduced latency is always a plus.
      • by dgatwood (11270)

        It's not like one gets cumulative latencies - the 230ms is constant in time.

        Actually, you do. Consider HTTP:

        • Client requests the page text.
        • (latency x)
        • Client gets the response and requests the JavaScript code.
        • (latency x)
        • Client gets the JavaScript code and runs it. The code requests several images.
        • (latency x)
        • The images are displayed.

        Total effective latency for the images is 3x. Latency is very much cumulative when one request depends on another. It is only non-cumulative when you are solely using the lin

  • by Anonymous Coward

    Oh awesome, now I can play on japanese quake servers!

  • Quake 3 on a 230ms connection would be awful. 170 might at least be playable.
  • Millisecond trading (Score:5, Interesting)

    by Michael Woodhams (112247) on Wednesday March 21, 2012 @12:30AM (#39423345) Journal

    Investing in shares for time spans of months is of general benefit to the economy, directing investment dollars to those best able to use them. Millisecond trading is of no benefit to anyone except millisecond traders, and any money they make is at the expense of people trying to do something productive. I propose that stock markets shift to a 'clock pulse' trading model: Trade bids for (e.g.) Apple are accumulated for (e.g.) 5 seconds and then all sales are resolved without regard for the order in which the bids arrived. This will cause no problems to real investors, but will rid us of the millisecond leaches.

    However, I am not experienced with the share market, so constructive criticism is welcome.

    • by 0olong (876791)
      What you propose is irrelevant. The stock markets are operated by the same people that benefit from HFT. They have no incentive to change. If you don't like that, make your own market place. Of course, to do so requires a lot of money and power, which you probably don't have unless you're one of said people to begin with.
      • by DarkOx (621550)

        Outside of the occasional "flash crash", which by the way have been mostly handled by the SEC reversing sales and pushing the reset button pretty effectively, I see HFT as mostly harmless. It might not add any value to the enterprises being traded but it makes the market better for investors not worse over all. The more orders out there at any time means the better shot an actual investor has at having their order filled when the want it.

        Liquidity in the market is a good thing. I don't have to wonder if

    • Re: (Score:2, Interesting)

      by Anonymous Coward

      Cross-exchange arbitrage strategies are typically performed by servers located as close the the midpoint of the cabling between them as possible. Perhaps it's time to speculate on Alaskan rack space.

    • Parasitic trading (Score:5, Interesting)

      by Anonymous Coward on Wednesday March 21, 2012 @12:57AM (#39423499)

      Parasitic trading is tolerated not desired. It diverts profit from investors into traders, reducing the number of investors in a market by reducing the profits they can make and thus reducing the capital available to companies. Fewer companies go to the stock market to obtain capital as a result.

      So yeh, you basically understood it correctly, however it has little to do with 'share' trading, rather derivatives.

      The derivatives market far outweighs the shares market these days. These are pure bets stuff like: "derivative X pays out k(Z-W) for each cent asset Z rises above (K+U+Y)/3.... ladies and gentlemen place your bets I will spin the wheel". It's a bookies pure bet.

      Unlike a proper bookies, Wallstreet pays out more money that it receives, so banks around the world place bets on these derivatives in order to make money. The banks and Wallstreet can afford to buy cables, it's pocket change since the underlying asset may only be a shopping mall worth $50 million, but the derivatives derived from that can be worth billions since it's a virtual asset with no real value beyond the fact it pays out a profit.

      In a good year (when they take more money than they pay out) Wallstreet awards themselves big fat bonus's, in a bad year, the Fed extends them more credit against smaller assets. So overall, because they pay out more than they take in, their borrowing leverage increases. Today it's something like 30:1 or more.

      The Fed says 'the loans were good we got all the money back', but that's a lie. They print money against 'Linden dollars', Wallstreet buys assets that pay out enough to cover the interest with that cash, Wallstreet borrows against those new assets, and pays back the money borrowed against the 'Linden dollars'. The Fed says 'hey look we got all our 'Linden loans' back', Wallstreet gets to own a real asset, everyone holding dollars has been silently robbed by inflation.

      But hey - faster internet! /rant

      • by tlhIngan (30335)

        Parasitic trading is tolerated not desired. It diverts profit from investors into traders, reducing the number of investors in a market by reducing the profits they can make and thus reducing the capital available to companies. Fewer companies go to the stock market to obtain capital as a result.

        Once a company IPOs, that's pretty much all the capital they get. And the bank handling the IPO pays the company to do so at a fixed price. Now, a company may choose to only sell say, 10% of itself, holding onto 90%

        • by Rich0 (548339)

          You don't need HFT to have market markers.

          If you think a stock is undervalued you buy some of it. If you think it is overvalued you sell some of it.

          In your example there are people selling at $6, and buying at $4. If I think the intrinsic value is $5 then I can put out buys at 4.50 and sells at 5.50 and be the first person anybody goes to to make a trade, and make a profit on every one (assuming the real value really is $5).

      • by jpmorgan (517966)

        Sigh, you have a very incorrect understanding of derivatives. First off,

        These are pure bets stuff like: "derivative X pays out k(Z-W) for each cent asset Z rises above (K+U+Y)/3

        No derivative works that way. Derivatives are, for the most part, options that allow you to make a trade in the future at a predetermined price today. For example, I can buy a put option which gives me the option of selling some shares 6 months from now at a preset price. Of course, it's not just shares. Futures are most heavily traded on commodities (wheat, oil) and currency (the US dollar, the euro, the yen).

        You're right that they can

      • by dj245 (732906)
        Unlike a proper bookies, Wallstreet pays out more money that it receives, so banks around the world place bets on these derivatives in order to make money.

        Maybe I don't follow this completely, but I take issue with this. Wall street is basically revenue neutral. For every winner there is a loser, and for every dollar in, that money has to come from somewhere because no wealth is created.

        You can argue that value is created, because you can buy something which you value but which somebody else doesn't
    • Investing in shares for time spans of months is of general benefit to the economy, directing investment dollars to those best able to use them. Millisecond trading is of no benefit to anyone except millisecond traders, and any money they make is at the expense of people trying to do something productive.

      With the sole exception of IPO's the only difference between holding a stock for a millisecond or for months is how long you hold the stock. In both cases you pay the guy you bought the stock from, and get

    • by RockDoctor (15477)

      Investing in shares for time spans of months is of general benefit to the economy,

      Many would argue that investing for periods of less than a [report cycle] is simply going to encourage [Management] to play to the short-term crowd.

      My employers have managed about 8% year-on-year growth for the 20 years that I've been an employee. But if we'd been a publicly-traded company, we'd probably have been shut down and lost the personnel asset several times in that period. 1.08^^20 = 4.7 ; good enough?

      No, probably n

  • by billcopc (196330) <vrillco@yahoo.com> on Wednesday March 21, 2012 @12:31AM (#39423347) Homepage

    "The latency drop will mainly benefit algorithmic stock market traders"

    In other words, these cables will help machines ruin the global economy.

    A part of me is kind-of glad they're speeding this up. We all know the system is destined to break, so the sooner that happens, the sooner people will wake the fuck up and demand change.

    • by Anonymous Coward

      Stock market doesn't mean shit about the economy. It's not a "good economy" when stocks go up--it's just inflation.

      • by gl4ss (559668)

        using hundreds of millions dollars worth of work time for it means shit for economy though..

        and probably the op was referring to stocks getting prices not based on their worth, but by algorithms which trade with other algorithms which trade based on what other people trade and not based on what the actual company worth is or is going to be.

        who gives a shit though? also automatic traders will have serves in japan if they're trading in japan so this(new cabling) isn't really going to matter at all to them. th

        • also automatic traders will have serves in japan if they're trading in japan so this(new cabling) isn't really going to matter at all to them

          The traders it matters to are those engaging in arbitrage.

          Many things (stocks, commodities, derivitives whatever) are traded on more than one market. Each market develops a price for the thing. Prices in the different markets are held reasonablly close by people engaging in arbitrage (buying in one market and selling in another). The lower you latency you have to BOTH markets the lower the risk that prices will shift to an unfavorable position while you are in the middle of executing your arbitrage.

  • by Compaqt (1758360) on Wednesday March 21, 2012 @12:32AM (#39423359) Homepage

    Surface of the water or of the seabed?

    I'd say you wouldn't need more than a few meters below the seabed.

    On the other hand, since the depth of the ocean may vary considerably, what sense does it even make to say they're burying it hundreds of meters below the surface of the water?

    That's like specifying underground land line depth in feet below the mesosphere.

  • Making repairs is a going to be major undertaking.

    Looking at the map, latency could be reduced further by routing via the North Pole. Of course that makes the troubles with laying and repairing the cables even worse.

  • Neutrinos (Score:5, Interesting)

    by mbone (558574) on Wednesday March 21, 2012 @12:48AM (#39423441)

    If you want to cut latency, communicate through the Earth with neutrinos [scientificamerican.com]. If we could just get the bit rate up some (from the current 0.1 bps), you could communicate to anywhere on Earth with a one way time of 40 milliseconds.

    • by bryan1945 (301828)

      If the 0.1bps is faster than light, what is the practical real-time bandwidth?

    • Re:Neutrinos (Score:4, Insightful)

      by Michael Woodhams (112247) on Wednesday March 21, 2012 @01:59AM (#39423787) Journal

      That would be from anywhere on Earth with a high intensity particle accelerator to anywhere on Earth with a huge particle detector buried hundreds of metres underground.

      • by timeOday (582209)
        Wow, I wonder if anybody is exploring this for communicating with or detecting submarines.
    • Re:Neutrinos (Score:5, Interesting)

      by erice (13380) on Wednesday March 21, 2012 @02:01AM (#39423795) Homepage

      That's true as long as it doesn't take any time to detect and decode a signal sent with neutrinos. Neutrinos are not electrons and trying to extract a signal from them is challenging enough that, in the near term, the computational latency would likely dwarf the transit time.

    • by InterGuru (50986)

      It's already been done [physicsworld.com].

      The first ever transmission of information using a beam of neutrinos has be achieved by physicists in the US. The demonstration is highly preliminary – it operates at less than 1 bit/s – and will require a lot of development before it can have any useful application. Nevertheless, the work proves a concept that physicists have been contemplating for years and that could ultimately be used in situations where other means of communications are not feasible.

      .

  • by hoggoth (414195) on Wednesday March 21, 2012 @12:49AM (#39423447) Journal

    The higher estimate of $1.5 billion is contingent on using Monster Cables.

  • The entire circumference of the globe is about 24000 miles, which takes light 128ms (in a vacuum). The article's claim that the current time to send a packet from London to Tokyo is 230ms therefore seems doubtful. In this time, light can go 42840 miles in a vacuum -- or nearly twice around the the entire distance around the world (24091 miles). Light in a fiber is about 35% slower, but this still leaves time for 37177 miles.
    • by Anonymous Coward
      Because internet traffic isn't converted or delayed while it goes around the world.
    • by ModernGeek (601932) on Wednesday March 21, 2012 @12:56AM (#39423491) Homepage
      Yeah, but the packets have to be routed, and that takes time, too.
    • by doshell (757915)

      You also have to account for the queueing delay at the routers, which are store-and-forward devices. That said, I really have no clue whether 230ms is a realistic number.

      • by Anonymous Coward

        If it travels through USA, it seems reasonable:

        $ ping visitfinland.ru
        PING visitfinland.ru (109.70.163.2): 56 data bytes
        64 bytes from 109.70.163.2: icmp_seq=0 ttl=55 time=199.910 ms
        64 bytes from 109.70.163.2: icmp_seq=1 ttl=55 time=198.688 ms
        64 bytes from 109.70.163.2: icmp_seq=2 ttl=55 time=203.657 ms
        64 bytes from 109.70.163.2: icmp_seq=3 ttl=55 time=202.524 ms
        64 bytes from 109.70.163.2: icmp_seq=4 ttl=55 time=202.258 ms
        64 bytes from 109.70.163.2: icmp_seq=5 ttl=55 time=201.571 ms
        64 bytes from 109.70.163.2:

      • by Sneeka2 (782894) on Wednesday March 21, 2012 @01:37AM (#39423695)

        I really have no clue whether 230ms is a realistic number.

        I currently get a 431ms Japan <> UK ping on a pretty mediocre Japanese ADSL line in the country side.
        So, yes, that's realistic.

        • by isorox (205688)

          I really have no clue whether 230ms is a realistic number.

          I currently get a 431ms Japan <> UK ping on a pretty mediocre Japanese ADSL line in the country side.
          So, yes, that's realistic.

          I imagine the route goes via the highly-congested south-east asia area, or possibly even via the U.S.

          London to Tokyo is about 6,000 miles via Sibera. Via undersea cables through the red sea it's nearer 12,000 miles. Via the states it's also about 12,000 miles.

          I haven't got any private lines to Tokyo, but my line to Singapore runs about 220ms IIRC.

    • You forgot amplifier latency and modem latency. Also packet acknowledgment.
    • by AHuxley (892839)
      Uk, many severs (US base), water, loops around natural and unnatural problems, Japan (US base), many severs.
      Even if the UK and Japan side offer real dedicated vs best effort, you still have to move around both regions before you hit pure new optical.
      Peering, telco deals can all send your packets for small or long loops before they get to fancy new projects.
    • by mr_walrus (410770)

      fibre optic repeaters (amplifiers) had huge latencies (relatively speaking.)

    • by gl4ss (559668)

      and distance between tokyo and london (short) is just 5900 miles.

      (but of course the cable doesn't go direct.. just saying that circumference of the globe isn't that useful to use in that)

    • Unfortunately fibers don't run in straight lines from place to place and there are processing/routing overheads too. Even so based on my own test it seems the authors of TFA are making the newbie mistake of confusing one way latency with round trip time (round trip time ~= 2x one way latency).

      Tracing route to www.jp [210.157.1.134]
      over a maximum of 30 hops:

      1 1 ms 1 ms 1 ms gw-umain.ee.umist.ac.uk [130.88.118.250]
      2 1 ms 1 ms 1 ms gw-rh.its.manchester.ac.uk [

  • by grelmar (1823402) on Wednesday March 21, 2012 @12:58AM (#39423507)
    There are so many internet applications where low latency is a win. VOIP, remote systems management, two-way graphical applications that for various reasons are location sensitive (there's more of these types of apps then you realize - think proprietary software that would be either illegal or economically dangerous to physically locate outside of NA or the Eurozone, including geophysical analysis software for mining/oil exploration, among other things)...

    There are lots of scientific applications where latency is critical.

    But oh, that would be difficult to discuss. Much easier to relate everything to a vilify-able application.

    Come on, for once, talk about the benefits of a mega infrastructure project.

    Oh, right... Slashdot. My bad. That's just not what we do here.
    • The only ones willing to pay a premium for the low latency are the HFT folks. I dont think, the masses who use VOIP, choose their ISP based on the ping time to UK (most go for the obvious bandwidth). I agree, it is a win for a lot of people, but it is paid for (atleast the initial costs) by High Frequency Traders.

      • I am slightly confused. Unless you want to do arbitrage between London Stock Exchange & Tokyo Stock Exchange, why not host your algorithmic trading servers at Tokyo? I think these days stock exchanges themselves offer hosting for lower ping times...

        Anyway, lower latency is always good, I don't really care if it's going to be used for HFT or not.

        --Coder
      • Tele-medicine.

        Or, more to the point: Tele-surgery.

        • by necro81 (917438)
          People talk about this as a potential killer app, but the simple fact remains that doing telesurgery with round trip communication times of a few hundred milliseconds is simply a non-starter. Surgeons work on hand-eye coordination and, to a lesser extent, tactile feedback. Introduce a quarter-second delay into that control loop and you either 1) royally screw up your position/velocity/force control or 2) maintain control, but at glacial speeds. In either case, it is doubtful that you would have surgical
  • No matter what they do to fix this, no matter what they do to reduce the latency I NEVER win any radio call-in shows. You would think this would help me be caller #25 but it never works. It's always busy or I just don't make that number :(

    • by Sneeka2 (782894)

      Well, soon you'll be able to try your luck on Arctic call-in shows. I'd say the chances should be pretty decent there.

  • "billions of people could wake up" well that's certainly a reason NOT to do this.
  • by bedouin (248624) on Wednesday March 21, 2012 @02:35AM (#39423931)

    See this [slate.com].

  • by dackroyd (468778) on Wednesday March 21, 2012 @04:37AM (#39424443) Homepage

    with a head office in the UK, I think this is awesome.

    Currently the packets between Oz and the UK either go through central Asia, where there is massive packet loss, or they go the long way round - across the Pacific, across the USA and then across the Atlantic.

    The new route will probably shave 40ms off the ping time from Oz to the UK as well as be pretty reliable - and also not subject to US data monitoring.

    • "...and also not subject to US data monitoring."

      Do you really imagine that the US and UK don't work together on this?

  • There is also the good side that this will bring serious bandwidth to places where dialup over satellite is currently the way to get a bit of Internet. Scarcely populated places in northern Canada and Alaska will appreciate the chances of a bit more bandwidth!
  • The latencies while downloading hentai and tentacle porn are crippling.
  • I was in South Africa in 2010 when one of their main cables got cut by a ship. It sucked. Internet got SLOOOOOWWWW

One small step for man, one giant stumble for mankind.

Working...