The optimal locations for HFT systems

AS forex traders arbitraging across exchanges attempt to eke out every possible advantage, they are reaching the point where they are being held back by the laws of physics.

For the HFT industry, the holy grail of zero latency eludes them. No matter how rapid trading software becomes at generating trades, the march towards zero latency is limited by the speed of light, as all data needs to be sent along a fibre-optic cable. Light propogation in a fibre-optic cable carries a constant latency of 5 microseconds per kilometre. As a result, traders are not just looking at the most advanced low-latency platforms but also the physical positioning of where their dealing takes place. “The location of trading servers in relation to the matching engines is paramount to pursuing latency sensitive trading strategies,” says Tim Bevan, director of global electronic trading services at Otkritie Capital. “When you have execution venues trying to shave off nano-seconds from their technology to compete with each other, line latency is likely to be the major cost component in terms of speed.” Bevan adds: “There’s no point building a Formula One car and trying to run it up a dirt track.”

And, much as Formula One teams focus on everything from the choreography of the pit-crews to achieve the perfect tire change to shaving off every unnecessary gram from the car itself, high frequency traders are looking for every opportunity to get the winning edge. According to Pierre-Francois Filet, chairman and co-founder of QuantHouse: “There is certainly a growing market demand for low-latency trading infrastructure within the FX market. Most firms have focused on addressing the three major technology areas to address the latency issue: network optimisation, low-latency connectivity and proximity.”

In a paper written by physicist Alexander Wissner-Gross and mathematician Cameron Freer – titled “Planetary-scale computing architectures for electronic trading” – the two propose ideal locations for the positioning of trades, depending on the financial instrument involved. And it’s not just a case of being equidistant between the two exchanges. If traders are seeking to mechanically arbitrage the difference in speed between one location and the other, then the ideal position moves towards the faster exchange. For example, when trading between New York and London, they identified an optimal position off the coast of Newfoundland. For obvious reasons, it is difficult to persuade traders to leave their Fulham and Upper East Side homes and set up shop in a dinghy in the middle of the Atlantic. But as Toby Corballis, chief executive of Rapid Addition points out: “Trading desks themselves are not necessarily moving in order to optimise latency advantage, but the siting of their algo boxes is critical. So while it is becoming increasingly important for the strategies themselves to be located in the right place, the location of the desk is becoming less so.”

Related articles