Latency Special: Will wireless trump fibre in the race to zero?
The world’s top firms have put their secretive but lucrative trading programmes on a fibre diet. Richard Irving finds out why.
Is fibre dead? Don’t groan – this is not some bad joke. So great is the seismic shift to wireless technologies in the ultra-low latency market, that some providers are seriously beginning to wonder what sort of future fibre can possibly hope to have in the do-or-die arms race to make share trading as fast as the speed of light.
And make no mistake, this is a race - as fast as a 100m sprint, as brutal as a back-to-back marathon, but above all, a race that no one can afford to lose.
The speed at which the ultra-low latency market has jumped from state-of-the-art optronics to a fifty-year old technology that struggles come rain or shine (or fog, or snow or high winds…) is perplexing.
That trading firms will turn their backs on ultra-reliable networks that might fail as little as once a year, to systems that can go down once or twice a day is all the more so.
But the fact that such firms will pay handsomely for the privilege – and to carry payloads that redefine the meaning of the concept mission critical, is just plain astounding. “If you had said to me a year ago that the ultra-low latency market would be obsessed with wireless, I would have said you were mad”, the head of marketing at one firm that specialises in the sector says. “If you had then gone on to suggest that trading firms would junk the state-of-the-art algorithms that have cost them tens of millions to develop, purely to make wireless work better for them, then I would have walked you to the sanatorium myself.”
Such is the latest twist in a market that never fails to surprise.
Back to the future
Driving this new push are a handful of secretive trading firms – no more than 12-15 – for whom speed is everything and cost means nothing. These firms, none of whom are household names, will willingly stump up as much as ten times the going rate for connectivity on ‘commoditised’ routes and considerably more in emerging areas. All they ask for in return, is that they be guaranteed the very fastest access to financial markets possible. These firms are often described as high frequency traders and it is widely assumed that they use highly sophisticated computer programmes to exploit tiny aberrations in the price of shares in different financial markets.
In truth, no one really knows how they make their money and a better description might be to call them electronic traders. One suggestion is that some firms might have cracked a way of trading into the “future” by effectively running rings around the trading engines of the stock exchanges themselves.
One firm claims to have found an example of a share that changed hands at a price that didn’t officially “exist” until some 190 milliseconds after the deal had been struck, though such evidence is hard to corroborate and it is not clear that regulators have taken the findings seriously.
All that anyone really knows for sure is that stock market trading engines work on a first-come, first-served basis and that these firms only profit if they are always at the head of the queue.
Behind this clique of trading firms is a wider universe of 50-60 banks and asset managers who would see themselves as very low-latency dependent, but who will not chase ultra fast connectivity at any price. And behind them are perhaps a few dozen trading firms who also need ultra-fast connectivity but will hold out for a better deal.
The total market for ultra-low latency services is considerably lower than many estimates suggest, which typically put the addressable number of customers in the low- to mid-thousands. But while the number of players is small, they are nevertheless determined to minimise risk and maximise profits by using the fastest networks that money can buy.
And if analysts at Rosenblatt, the Wall Street broker, are right these speed-junkies made about $1.8 billion in 2012. Those estimates are supported by academics at Princeton University, who reckon that electronic traders collectively made around $23 million last August in one single security alone – the S&P 500 futures contract – and all this, in one of the quietest months of the year when any self-respecting high-roller is holidaying in the Hamptons.
Like his peers, Hugh Cumberland, a business development manager at Colt is surprised at the speed at which this community of electronic trading firms has embraced wireless technology, with all its faults. As recently as six months ago, Colt had not given serious consideration to adapting microwave or milliwave systems for financial customers. Today, Cumberland is putting a new wireless link between London and Frankfurt – the world’s busiest trading route - through a final round of tests before going live over the course of the next few weeks.
Late last summer, he says, Colt began to pick up murmurings that engineers were close to putting the landward sides of the London to New York link onto microwave frequencies. In the end Perseus Telecom confirmed that it had done precisely that, knocking close to 4 milliseconds (mS) off the current record (which at the time was held by Global Crossing, now part of Level 3).
“I thought something similar might happen to the London-Frankfurt route. But in the end we found we could do the whole link with microwave, allowing us to reduce latency by 40%”. Cumberland says.
In keeping with other ultra-fast network providers, Colt is no longer willing to disclose exactly how fast it can ping data across its ultra-low latency networks for fear of setting a benchmark for rivals to beat. However, ahead of the switch, Colt is understood to have been able to do the round trip in just under 9 mS, suggesting a new record of around 5 to 6 mS.
That would put Colt’s new London-Frankfurt offering to within 1 mS of the speed of light itself – the ultimate goal in the so-called “race to zero”. A millisecond is one one thousandth of a second: to put that kind of speed into some context, it takes about 300 mS for the human eye to blink.
So successful has the roll out been, that Colt, which has long prided itself on its connections with the financial services industry (no pun intended), is now building out wireless links to other key financial routes. “We’ve had no surprises or disappointments so far which has led us to start assessing the potential customer demand for microwave links elsewhere”, he adds.
When it comes to network architecture, speed is just one part of the story - capacity and reliability also play crucial roles. But the problem with a wireless network is that, like a Formula 1 racing car, it’s built for speed, not capacity. And when it comes to reliability, it’s not so much a title-winning Renault, as a bottom-of-the-table Cosworth.
Downsizing big data
There are two main ways to deliver data wirelessly: via millimetre frequencies, which can only travel very short distances before they need “boosting” but which can carry up to 1 Gigabyte per second (Gbs) of data, and microwave, which can travel further but can support as little as 150 Megabytes per second (Mbs) of digital information.
Neither should be particularly attractive for high frequency traders, who trawl vast banks of financial information across different asset types and around different market places in search of fleeting anomalies. If ‘Big Data’ were ever to need a poster boy, then the electronic trading community would be a shoe-in.
Or so you might think. In reality, the high-octane world of electronic trading is governed by just one rule: speed trumps all. And so one of the key reasons that wireless has taken off so quickly, is because traders have re-tooled their strategies to adapt to the lower capacities that wireless networks can support.
“The guys that are making the money in electronic trading these days are the antithesis of Big Data – they are really, really fast Little Data”, explains Mike Persico, chief executive of Anova Technologies, a Chicago-based firm that provides connectivity to specialist trading firms.
“The smart guys have already worked out how to shed bandwidth because that way, they can get themselves on a milliwave architecture for not much more money than it would cost them to run a 10Gbps feed on a fibre route.”
Jock Percy, chief executive of Perseus Telecom, agrees: “A lot of trading algorithms were supersized. As a result of wireless technology, they’ve been put on the treadmill. Trading engines have got a lot smarter – they don’t need multiple wavelengths of capacity anymore.”
Official figures back this up. Last year, the average daily number of shares traded in all US stocks fell by 15% to around 6.8 billion against 2011, while the percentage of those volumes directly attributable to high frequency trading fell from 55% to 51%. At its height, US regulators believed that anything up to 80% of all dealing volumes in America were as a direct result of high frequency trading.
Race to zero becomes a race for space
The challenges posed to wireless technologies by the weather are as well documented as they are diverse. You name it, wireless plays up in it: snow, fog, bright sunlight… the list goes on. There are other obstacles too, such as large expanses of water and even trickier problems such as military installations and airports. Even flocks of birds have been know to attenuate microwave signals. Millimetre wavelengths are even more susceptible to interference because the signal is weaker to begin with.
“Millilmetre wavelengths hate, hate, hate rain”, says Persico. “If you try to set up a network without protection against rain you’re going to be down 70% of the time.”
Then there is the whole issue of line of sight. Wireless signals need to be refreshed at regular interviews, which means operators need a string of radios and antennae on towers and high buildings right along the route as well as a license from the Federal Communications Commission to use a specific radio frequency. Space on such towers and the spectrum to transmit signals from them are getting increasingly hard to come by.
As David Selby, vice president of Product and Strategy at euNetworks explains, building a wireless network from scratch at short notice is a very complex job: “You have to juggle what type of equipment you can source for the frequencies that are still available and based on where exactly you can place the equipment. It’s a complex implementation process that takes several quarters and a lot of engineering to accomplish."
Undoubtedly, the challenge is to make wireless technology a “five-nine’s” proposition – that is, make it reliable 99.999% of the time.
If microwave, by dint of the severe capacity restraints it places on the user, is Betamax and millimeter is VHS, then Persico believes he has come up with the network equivalent of blue ray DVD: “We have developed high availability gear – a hybrid system which can switch from millimeter to microwave if a storm sweeps in.” The system is particularly well suited to the microclimate of America’s East coast, Persico says: “We can now stay up in pretty much all weather conditions.”
Still, the issue of reliability is perhaps the reason why fibre will retain an active role in most ultra-low latency networks. The current expectation is that electronic firms will switch to low bandwidth wireless for execution-only algorithms and stick with ultra fast fibre for the more brainy analytics. Moreover, some of the more secretive operations will doubtless be nervous of the potential security breaches that might arise in wireless connectivity, preferring the relative safety of sub-terrestrial fibre.
“You have to be very cautious as to the investments you make for the future versus the size of the addressable market”, warns euNetworks’s Selby. “The ultra-low latency market has come out of nowhere and grown very quickly, but its absolute size is limited. I don’t necessarily think there is a line in the sand where, once you take that first step, you have to commit fully to wireless at the expense of all else”, says Selby. “It’s not at all clear, even now, whether everyone in the electronic community will jump to wireless - some customers in the market may never do so at all.”
Colt’s Cumberland also sees a rosy future for fibre. Clearly, he says, providers are going to find it imperative to maintain ultra fast fibre routes well into the future: “Our challenge is to leverage our existing investment in fibre while adapting to customer requirements that want us to make the best possible use out of wireless technologies.”
One way is to start cross marketing both wireless and fibre alternatives: “We’re in the unique position of being able to offer a package of fast fibre and fast microwave and that makes for an interesting sales strategy.”
Whither Spread Networks?
One consequence of this remarkably rapid shift to wireless technology is that it appears to leave Spread Networks, the father of today’s multi-billion dollar electronic trading market, out on a limb.
But Spread, let it not be forgotten, pioneered low latency connectivity by slicing more than 100 miles and 3 mS off the route from New York to Chicago with a new $300 million network laid in secret and timed to harness the explosion in high frequency trading back in 2010.
Spread declined to comment for this article. That may mean nothing, or it may mean that the master of subterfuge is once again back in stealth mode.
“The greatest mystery to me, is not so much what role fibre plays in the new paradigm, but rather what role Spread plans to play in it”, says the chief executive of a rival network provider.
Spread saw the potential of ultra-fast trading long before anyone else. To suppose that it never saw wireless coming or indeed that it wouldn’t move aggressively to protect its asset, is to gravely underestimate the company’s management.
One option might be to buy a smaller wireless operator, though Spread’s record suggests it has little appetite for mergers and acquisitions and would prefer to build out its own technology.
Alternatively, it might be looking at a new technology such as hollow core fibre. As its name suggests, the middle of a length of hollow core fibre is made up of a vacuum, which allows light to travel through the fibre at about the same speed as it would over a wireless system. But because it is laid underground, it is free from interference and is thus extremely reliable.
The new-style fibre is still in early stage testing and engineers say that it is at least a decade away from mainstream production. And while it undoubtedly has potential, some insiders point out that any hollow-core fibre network would have to be laid from scratch and in a very straight line to compete with wireless equivalents in the ultra-low latency market.