High-frequency traders use super-fast computers to trade automatically hundreds of times a second

The advantages of faster connectivity are rarely questioned these days, with consumers told it will improve their access to lifestyle and content services and enterprise users focused on improved productivity. Often the touted benefits are, to a degree, abstract. But in the world of high-frequency financial trading (HFT), the benefits are very measurable and very significant.

High-frequency traders use super-fast computers to trade automatically hundreds of times a second, often arbitraging tiny price differences between stocks listed on different venues. Added up over the course of time, the combined value of all those tiny profitable trades quickly builds into vast mountains of cash.

Scarcely a day goes by without another IT vendor, bank or system operator rolling out a new low-latency link. In the world of securities trading, milliseconds can literally translate into millions of dollars. At the high end of the spectrum, a time difference of even a fraction of a millisecond can make the difference between the winner and the loser in a financial transaction.

Experts report that shaving a single millisecond from every trade can be worth upwards of $100m a year to a large trader. And the traders operating in this space are prepared to pay a premium.

Serving the needs of these low-latency market participants, a range of high-speed broadband links have emerged across the globe. Deep sea cables, such as the 40Gbps Asia Submarine-cable Express, which opened at the end of August, cross the oceans between Singapore, Hong Kong and Japan.

Firms such as Hibernia Atlantic claim to offer the fastest possible connection between North America and Europe via the UK. The SunGard Global Network offers securities firms connectivity and trading tools for markets around the world.

“The importance of being first is so great that HFT firms will do almost anything if it will increase their speeds,” says Hugh Cumberland, business development manager, consulting and solutions at communications provider Colt. “If I take one millisecond off their round-trip delay, the customer may find the cost of that service virtually irrelevant. Most HFT strategies are based on being first or equal, so if the customer is getting there first every time, the return they can make is almost unlimited—at least until the competition catches up.”

In tandem with the increasing speed of connectivity networks, stock exchanges and securities trading venues have also increased the speed of their trading engines dramatically. As recently as five years ago, the time delays between sending orders and receiving confirmation of a match on Japan’s Tokyo Stock Exchange, one of the developed world’s champion exchanges, was measured in whole seconds. Now, exchanges measure latencies in milliseconds, while the most cutting-edge technology providers talk in microseconds.

Not all high-speed traders arbitrage price differences. HFT encompasses a whole range of strategies, some of which involve market making; posting quotes at the bid and the offer, thereby helping to provide liquidity in the hope of making a profit. But regardless of the definition, little dispute exists that HFTs represent both a substantial portion of equity market volumes—up to 70 per cent in the US, and between 40-50 per cent in Europe—and an equally important customer segment of exchanges, which have become ever-more attuned to the need to attract HFT flows to drive up liquidity levels and bolster revenues.

Put simply, a large factor in the drive to increase speed at the exchange level has been the desire by top exchange executives to capitalise on the HFT dollar.

Naturally, connectivity providers have been keen to establish close relationships with the exchange venues, while exchanges have sought to make themselves as accessible as possible. The result has been phenomenal growth in co-location services, in which vendors and market participants locate their trading infrastructure as close as possible to the exchange matching engine.

Earlier this month, market data and trading infrastructure services provider MarketPrizm opened a co-location environment at the Tokyo Stock Exchange, to provide users with market access services including data, order routing, managed hosting and networking in the TSE data centre. It includes connectivity for equities, exchange- traded funds and derivatives. Users can also gain access to data from other Asia Pacific venues via MarketPrizm fed to customers inside the TSE data centre. The idea is to make it easier for latency-sensitive investors to access the market quickly and cost-effectively. “Our solution lowers the cost of entry into Japan and enables overseas customers to trade without a local office in Japan,” says Tanuja Randery, CEO at MarketPrizm.

“We can provide financial firms a fully managed market access solution with the best latency in the market for accessing the TSE. We can connect the buy-side with brokers for execution inside co-location and we can provide the brokers with a faster on-boarding program for their customers.”

Links such as the MarketPrizm connection are widespread across the globe. Brazilian exchange BM&F Bovespa is particularly notable for the emphasis it has placed on upgrading its own trading platform and providing co-location—a service that has been enthusiastically embraced by foreign investors keen to access Latin America’s biggest market. In August, they accounted for 40.02 per cent of total trading volumes in the Bovespa equities segment. The same month, the exchange’s co-location services registered 198,057 trades, worth $1.28bn, on the exchange’s Bovespa equities market—up from 118,033 co-located trades ($1.01bn) in July.

The use of co-location is not limited to equities, either. Brazil’s market for derivatives is huge; the exchange already has a routing agreement with US derivatives giant the CME Group that allows US investors to access the Brazilian BM&F derivatives market through their existing connections. But increasingly, trading firms are co-locating for derivatives trading at the exchange in Sao Paulo; the number of co-located derivatives contracts traded at BM&F increased from 8,066,765 in July to 8,915,606 in August. Foreign investors represented 25.37 per cent of activity on the bourse’s derivatives market.

Third-party data centre services providers have also established a viable niche for themselves, providing an environment for banks to interact with exchanges and technology vendors, all at blistering speeds many times faster than the blink of a human eye. The arrangement works to the benefit of all parties, according to those who operate the data centres.

“Co-locating to a third-party data centre provider brings benefits to all stakeholders involved in financial trading,” says Patrick Lastenet, director of marketing and business development for financial services at data centre services provider Interxion. “For distribution processes, financial institutions typically first receive the data and then have to redistribute it to their own customers down the line. In a co-location facility, however, customers can locate within the same data centre as the distributing firm and can cross connect, presenting opportunities for instant distribution and significant time saving. Furthermore, sharing a facility with other market participants can help create a hub topology that interconnects within itself, increasing firms’ reach within the data centre and encouraging a reduced time to market for acquiring new customers.”

According to Lastenet, it is in financial firms’ best interest to use a virtualised IT environment such as a data centre to cope with increasing levels of data and keep pace with real-time transaction records. With virtualised server farms, companies can pull more computing resources to complete the tasks without labouring their IT systems and limited physical capacity, he explains. In addition, since additional virtual machines can be spun up or down at will, firms can adapt to high-data volumes more easily than continually adding more physical boxes.

“Banks have traditionally built in house data centre infrastructures for security, compliance and reliability reasons,” he says. “Now regulatory compliance is placing a significant strain on financial institutions’ IT infrastructure, as they have to evaluate and cope with exponential increases in real-time data. With MiFID II set to go live in the near future, further real-time requirements will only amplify the need for high performance analysis of big data sets. In the face of this challenge many banks are now adopting the use of a third-party data centre for non-core business processes.”

From warp speed, to ridiculous speed, and finally into ludicrous speed...

In the cult 1987 comedy film Space Balls, which parodies the Star Wars trilogy, the nefarious antagonist, known as Dark Helmet, orders his ship to accelerate from warp speed, to ridiculous speed, and finally into ludicrous speed, with disastrous consequences for himself and his ship. So could today’s high-speed networks be in danger of flying too close to the sun, and ultimately getting their wings burnt? Fraser Bell, managing director at specialist low-latency telecoms operator BSO Network Solutions, Networks, believes the key factors are the fibre on the ground, the speed of light, and geography. BSO Network Solutions operates a London to Hong Kong link that reports 176 milliseconds latency from point to point.

The firm has also established a low-latency route into Madrid, which cut the latency by 0.5 milliseconds versus the best existing route to that market—enough of a benefit for US and European HFT firms to make it worth their while to connect, according to Bell.

“The opportunities exist to continually lower latency,” he says. “But the economic cost to build the infrastructure may not be worthwhile in every case. For example, only a very small subset of investors is willing to pay an extreme amount to save five milliseconds on a transatlantic link between markets that are already relatively well-served by low-latency technology. At BSO Network Solutions, we try to focus on emerging markets where we can add the most value at the lowest cost.”

Different providers disagree about exactly where the cut-off between justifi able cost and dangerous risk lies. Hibernia Atlantic, a provider of sub-sea cables that operates a link between North America and Europe via the UK, has invested millions of dollars in building sub-sea cables that connect the two continents. In total, the company operates 24,000 kilometres of fi bre network, representing just over half the circumference of the globe (40,075km).

According to Cumberland at Colt, the most direct route is always the fastest —yet many of the existing connections are surprisingly indirect, leapfrogging between major urban centres instead of taking the fastest possible route between two points. As the speed of connectivity approaches zero, the laws of physics start to become a problem. Andrew Bach, former head of network services at NYSE Euronext, famously proclaimed last year that the speed of light is “getting annoying” in the context of high-speed financial securities trading.

The speed of light varies based on the material used to build the connection—an effect that is measured on a scale called the refractive index. The fastest (Earthbound) medium is air. Travelling through a ground based cable is slower; light travelling through glass is only two-thirds as fast as light travelling through air. Changing the medium can help—for example, some firms are working on using hollow fibre and such a route already exists between Chicago and New York. However, some industry participants are sceptical about the long term prospects for increasing speeds purely through fixed lines.

“If it doesn’t incur a huge cost, the customer can gain a competitive advantage that goes beyond clearing and executing the trade quicker,” says Bell. “HFTs, banks, insurance firms and so on can push their business applications, such as Outlook, over the same system, which will allow all employees in the company to go faster.

But fibre can only go so far. The fastest route between London and Moscow is currently offering 40 milliseconds; last year it was 50. I suspect that 35 milliseconds is as good as it’s going to get using fixed ground-based lines.”

If there is a finite end to the ability of terrestrial and undersea cables to reduce latency, then developers may have to focus on making faster software and applications instead

If there is a finite end to the ability of terrestrial and undersea cables to reduce latency, then developers may have to focus on making faster software and applications instead. Already, companies such as Corvil have built entire business lines devoted to latency management. The firm’s flagship product is CorvilNet, a latency management system that provides monitoring, troubleshooting and reporting for financial trading environments. Trading firms invest in FPGAs to make their hardware faster. HFT operators try to reduce latency at the application layer and across all areas, not just the connectivity. But there may be another solution: wireless technology.

“Microwave is the fastest possible medium we have at the moment,” says Cumberland. “A few years ago, people would have scoffed at the idea of using it for communications such as sending information about financial transactions. Now, that has changed.”

While potentially faster, wireless connections are also more problematic for financial services firms. Wireless transmissions are more susceptible to weather conditions, storms and other interference that does not affect fixed terrestrial lines. The less predictable nature of wireless connectivity could be a serious source of concern to banks, which rely on predicting and controlling risk as a core part of their business activity. Yet Cumberland downplays such fears, pointing to capacity and power issues as more serious challenges to be overcome before the technology can gain widespread acceptance.

“Storms and environmental interference are not as serious a problem for wireless technology as you might think,” he said. “The real problem is bandwidth. We can send 1.6 terabytes per second in a cable. In a microwave, we have less bandwidth available. There is a millimetre wave in development, which can provide better bandwidth, but it requires huge amounts of power and therefore has less range.”

Whatever the solution that emerges, it is a safe bet that high-speed traders will continue to push the boundaries of technology for the smallest advantage. With former Goldman Sachs trader Anton Kreil poised to make the world’s first trade from space in 2014, the possibilities seem virtually endless.


Post your comment

@telecoms