Latency Special: How high frequency trading is driving low latency
06 September 2012 | Richard Irving
Time is money – especially in the cut-throat world of high frequency trading. In an increasingly savage arms race to slash connectivity speeds, an old technology is suddenly taking centre stage. Richard Irving reports.
An apocryphal tale is circulating among the rapidly growing circle of low-latency providers: it involves a large trading institution that was seeking 100G of capacity on a super-fast connection between New York and Chicago.
Urged to move quickly amid signs of strong demand, the bank retorted with a strange request: how much capacity was still available, it asked matter-of-factly.
The next day it couriered a cheque over to the provider for almost 100 times the original quote, snapping up close to 10G of ultra-low latency bandwidth in all. The bank stumped up the cash, so the story goes, not to safeguard its own future capacity needs but to squeeze rivals out of the market.
Then there is the case of another trading house that supposedly paid a network operator a $1 million bonus just to remove a loop of fibre nestling under a manhole cover a few yards from the New York Stock Exchange.
The excess fibre was reputedly slowing down the performance of its network so much that the bank’s highly prized "black box" trading system was failing to work properly.
But the most bizarre story of all is that of a California-based former rocket scientist who has hooked up with a veteran US Marine to dust off a decades-old technology that is poised to smash the current record for super-fast connectivity.
In so doing, it could reshape the ultra-low latency landscape and raise serious question marks over the future of thousands of miles of fibre currently linking New York and Chicago, two of the world’s largest financial hubs.
Wall Street’s dream
Welcome to the world of high frequency trading, which nestles on the wild frontiers of modern day telecommunications and which gorges itself not so much on the need for speed, but rather the greed for speed. It’s a world where tens of thousands of buy and sell orders can zip from one stock market to another in, quite literally, the blink of an eye and where a single millisecond advantage in connectivity can be worth upwards of $100 million a year in trading profits.
In its simplest form, high frequency trading is a bit like snatching £20 notes off the pavement – the first person to the money gets to pocket the cash. Banks use highly sophisticated computer algorithms to identify and then exploit minor aberrations in the price of similar stocks and bonds in differing markets across the world. It’s a high-volume low-risk business that accounts for anything from 50-80% of total volumes on some exchanges. But it relies on speed – only the first bank to spot a mis-match can profit from it.
Ultra-fast connectivity between markets is therefore of the essence. According to Tabb Consultancy, a US analyst, banks will spend around $2.3 billion on external networks this year as they invest to stay ahead of the competition with ever faster trading circuitry. Time is money and secrecy is paramount, which is why many of the stories that do the rounds of the publicity-shy financial community are difficult to verify.
But Bob Meade, a graduate in theoretical physics from Harvard and a former star trader at Ronin Capital, the US hedge fund – and his partner, Jim Sullivan, a former first lieutenant in the US Marine Corps with specialist knowledge of combat communications – are both very real indeed.
The pair’s venture, McKay Brothers, is currently building a microwave-based ultra -low latency network between Chicago and New Jersey that threatens to knock several milliseconds (ms) off the current fastest fibre route. That route, itself constructed under a thick veil of secrecy in 2010, is owned by Spread Networks and currently clocks in at around 13.10ms there and back. That’s about one tenth of the speed that it takes for the human brain to process a thought.
McKay, which is self-financing the network build, is using a microwave radio platform specifically designed for ultra-low latency applications by Aviat Networks. The firm, which plans to co-locate its microwave dishes on existing towers, says it has already cleared Federal Communications Commission (FCC) approval and aims to go live in the near future.
It will need to. According to FCC filings, there are around a dozen rivals currently looking to build wireless links between Chicago and New York, including efforts by Renaissance Electronics, a specialist manufacturer of microwave systems based in Harvard, Massachusetts.
McKay won’t say exactly how fast its new link will be. But it will need to challenge the 9ms barrier if it is to compete with another leading low latency microwave network provider, Windy Apple Technologies. New York-based Windy Apple, the self-styled pioneer of low latency microwave network design, launched what is widely thought to be the first fibre-beating link in 2010 for an unnamed finance house. At the time, the firm was gagged from either disclosing the speeds it was capable of achieving or the identity of its client. Those restrictions are now lapsing and the operator is looking to sign up new clients to its Chicago - New York link and to roll out a new network that could eventually extend from New York via Toronto to Asia.
Windy Apple is already thought to have breached the 10ms floor but it is the comments of yet another wireless competitor that will strike fear into the hearts of fibre network providers the world over: "We are already at 8.6ms round trip between New York and Chicago and with further enhancements we expect to get within 0.5ms of the speed of light", says Sal Benti, chairman and founder of NeXXcom Wireless.
In the so-called "race to zero" – the fiercely competitive dash to drive down network speeds – "zero" is quite simply the speed of light, or the shortest theoretical time that it could possibly take for a chunk of data to travel from point A to B. Physicists estimate that it would take around 7.78ms for light to travel in a straight line from New York to Chicago. If Benti can breach 8.5ms – or even the 8ms barrier, he will become the Usain Bolt of the low latency world: unbeatable, unstoppable and undisputed winner of telecoms’ greatest race.
Benti is winning some influential friends among the fibre community, not least Hunter Newby, chief executive of Allied Fiber and founder of the Dark Fiber Community: "In the ultra-low latency arena, microwave technology offers speed advantages with sufficient bandwidth to be most effective for certain high frequency trading applications. I suspect that as traders learn more of NeXXcom’s technology, they will begin proliferating these networks around the globe in the not too distant future", Newby says.
Poised for microwave?
Next in Benti’s sights are the subsea routes. An effort spearheaded by Jay Lawrence, NeXXcom’s top engineer will soon see the firm deploy millimetre wave technologies to hook up various landing stations on either side of the Atlantic with key financial centres in Europe and the US.
It is perhaps a little ironic that an ageing technology – MCI, lest it be forgotten, was using long-haul microwave networks way back in the 1970s – should be driving innovation in the world’s financial markets. But it should not be surprising. "The simple truth of the matter is that we are very close to the limits of what we can do with fibre", says Hugh Cumberland, a London-based financial services consultant with Colt. "Within a couple of years, every low latency route will have some sort of non-fibre element to it wherever technically possible and I would like to guess that it will be microwave."
So far, the switch to microwave technologies in the low latency field has been largely confined to North America, but a similar shift in European network design is inevitable: "If a large European player takes a big chunk off latency between, say London and Frankfurt, by knitting a microwave piece into the network, then everyone else will have no option but to jump", concedes Cumberland. "You won’t have a choice if you want to remain a leading player in the low latency market."
That day will soon be upon the fibre community. NeXXcom is close to completing a wireless link between London and Frankfurt, the busiest securities trading route in the world, that should breach the 5ms barrier for the first time. The route will almost halve current commercial speeds and usher in a new phase in European low latency network design.
This will pose complex challenges for many network operators: how much stomach will they have to continue ploughing investment dollars into fibre knowing that the world is poised to jump to a microwave topology?
"If you’re not the fastest network on the street, then you won’t hang on to the top tier of customers. Financial services companies will need to look for the best combination of technologies that keep them at the front of the game", explains Cumberland. That is likely to entail working to shorten fibre routes, bolster bandwidth performance and seek improvements to the equipment at either end of the cable, he explains.
Meanwhile, adds Erick Contag, chief operating officer of GlobeNet, the law of diminishing returns will hang over fibre routes: "The model on which ultra-low latency networks are financed and built relies chiefly on the premise that your customers do not switch to a rival technology or a new lower latency network before you reach your break-even point. That assumption could well be tested if a microwave operator delivers a meaningful improvement in performance over a fibre-based terrestrial network."
Light travels through air about 30% faster than it can through a glass fibre, so it follows that traders using microwave technology will pip their rivals on fibre routes to the most profitable trading opportunities. NeXXcom, for example, estimates that its microwave solution can knock around 5ms off the fastest fibre link from New York to Chicago.
And if the laws of physics play heavily into the hands of operators such as Benti, so too, do those of economics. NeXXcom reportedly spent just $6 million building its first Chicago/New York route. Analysts estimate that Spread Networks, meanwhile, spent around $300 million on its low latency route – a Herculean effort that lopped 100 miles and 3ms off the previous shortest route by laying a one-inch cable through Pennsylvania’s Allegheny Mountain range.
But there is a pay-off. For one thing, microwaves can only travel by line of sight, so a network is only as good as the next cell tower that it can "see". On average, that means finding a cell tower every 17 miles or so on which to put a microwave dish (NeXXcom has less than 30 on its bellwether route). And while engineers have managed to establish a microwave connection spanning a 360km-plus stretch of the Red Sea, a trans-Pacific or transatlantic network looks some way off. Neither is there a surfeit of cell towers on which to place dishes.
Moreover, some millimetre wave technologies do not travel well through moisture-laden air and are particularly prone to disruption in heavy rain, snow and even fog. Mountainous regions are tricky to circumnavigate and interference issues require network planners to steer well clear of airports and certain military installations.
But perhaps more importantly, they struggle to carry as much information: at the top rates of throughput of around 1G, wireless data rates are less than one tenth that currently available on standard fibre-optic solutions.
Going for gold
This last point is particularly crucial. As Jock Percy, chief executive of Perseus Telecom, explains: If the network is only being asked to handle straightforward buy or sell orders, then capacity constraints are unlikely to be much of an issue.
But if vast pools of market data are pulsing through the network in order to feed the decision-making algorithms that drive many high frequency trading computers, then bandwidth constraints could become a factor: "If you’re trading a wide array of financial instruments in several diverse geographical locations, then you could be running an extremely bandwidth-thirsty application."
Percy is a gutsy competitor in the race to zero. This spring, his company pre-empted the long-awaited launch of Hibernia Atlantic’s $300 million transatlantic submarine cable by announcing that it had re-engineered the land-side circuitry on Flag Atlantic’s northern cable to bring latency from Lands End, England to Long Island, New York, under 60ms round-trip for the first time.
Hibernia had already pledged to slice 5.2ms off the London-New York route when it announced plans to build the Express submarine cable, which is due to go live next year, but the initiative currently lies with Percy: "There’s no silver medal in this race– you either offer customers the fastest link, or you accept that you are a commoditised network option in a mature market", he says.
Nowhere is this more pertinent, he adds, than in the transatlantic market, which is currently running at less than one fifth of total available capacity and where the opportunities for charging a low-latency premium are rapidly shrinking.
Even capacity constraints are unlikely to keep wireless providers at bay for long: NeXXcom’s Benti predicts that radio waves will soon easily cope with bandwidth capacities of 1G and there are clear signs that trading firms themselves are simplifying their black-box systems in order to run programmes over lower capacity waves than through fatter, but slower fibre pipes.
If the outlook for fibre providers in the low latency market looks a little unsettled, there is at least some good news. Premiums for low latency links are holding up well and new markets are coming on stream.
Moreover, new European regulations due to come into force next January will require that more financial instruments be traded on recognised exchanges rather than privately between banks, offering more markets for network operators to tap into.
In Europe, new opportunities are emerging in Milan, Madrid, and to a lesser extent Zurich while elsewhere the burgeoning markets of South America look promising.
As Colt’s Cumberland concludes, it is the nature of financial markets to invent new markets and new asset classes: "Our task is to provide the connectivity to facilitate trade between them."
Is the race to zero winnable?
The challenge with serving the high frequency trading community, quips one veteran telecoms executive, is that you have to be in it for the very long term. The problem is that your customers most definitely aren’t.
The universe of potential top-flight buyers in the low latency market is relatively small – probably in the mid-hundreds, although many players look to take two or even three entirely separate routes in order to ensure diversity in the event of an outage. And while they will pay handsomely for that connectivity, firms will migrate to faster networks in a heartbeat.
Moreover, uncertainty hangs heavy over the sector, not least from regulators looking to clamp down on what they see as a systemic threat to market stability. The near-collapse of Knight Trading, a New York trading firm that sent the US stock market temporarily into free fall in August after the installation of a new black box programme went awry, will further fuel such uncertainties.
New technologies will also play their part. Some physicists, for example, believe that it might one day be possible to harness neutrinos in the search for ever-faster trading speeds. Neutrinos are subatomic particles that can literally travel through the centre of the earth unimpeded. If they could be encoded to carry buy or sell orders, they could cut the latency from London to New York (currently 60ms) to around 10ms.
Network providers who succeed in reinventing their low latency solutions to businesses outside the financial sector will clearly edge ahead of competitors.
While it is certainly the case that no other enterprise vertical is quite so obsessed with latency as to seek service level guarantees down to fractions of a millisecond, the world is becoming a more latency-sensitive place.
But while many potential customers might be interested in low latency, they rarely prove willing to pay a premium for it and they certainly will baulk at the pricing structures that banks are willing to accept.
The online gaming industry and the gambling sectors are tentatively exploring low latency applications while there are very clear synergies with the broadcast media sector, where the continuing shift to high definition TV is fuelling an effort to drive down latency.
Perhaps the ultimate winners in the low latency dash to zero, though, will be those network providers who take the lessons they have learned in developing the hugely successful ultra-low latency market, on to the next niche network opportunity.